PDA

View Full Version : KenPom 2019



dukelion
10-21-2018, 12:50 AM
Didn't see a 2019 Dork Poll thread so feel free to start one.

Anyways.....KenPom has us at #2 and the ACC heavily repped in the top 25.

And what's up with Nevada?

https://kenpom.com

Kedsy
10-21-2018, 01:05 AM
And what's up with Nevada?

https://kenpom.com

Top 25 and Sweet 16 team returns 5 of top 6 minute-getters from last season and brings in a top 20 recruit. #6 might be a bit high, but you wouldn't flinch if they were in the #10 to #15 range, would you?

HereBeforeCoachK
10-21-2018, 08:54 AM
Top 25 and Sweet 16 team returns 5 of top 6 minute-getters from last season and brings in a top 20 recruit. #6 might be a bit high, but you wouldn't flinch if they were in the #10 to #15 range, would you?

Yep, Nevada's real......

mgtr
10-21-2018, 08:55 AM
And the PAC-12 fell out of bed! Amazing - there is Oregon and then you have to look in the 40s for them. What would Bill Walton say?

Troublemaker
10-21-2018, 09:22 AM
Kansas' projection is somewhat interesting. While many places have them #1, here kenpom is projecting dominance from the Jawhawks. #2 offensive team, #1 defensive team. The difference between their projected efficiency margin and the #2 team Duke's is approximately the difference between Duke and the #10 team WVU, or the difference between WVU and the #26 team Texas Tech.

rasputin
10-22-2018, 12:01 PM
And the PAC-12 fell out of bed! Amazing - there is Oregon and then you have to look in the 40s for them. What would Bill Walton say?

Walton would say a lot, but not about the game he's broadcasting.

Wahoo2000
10-22-2018, 01:17 PM
And the PAC-12 fell out of bed! Amazing - there is Oregon and then you have to look in the 40s for them. What would Bill Walton say?

Not impossible that some mid-majors could send more teams to the tourney than the "conference of champions".

johnb
10-22-2018, 01:28 PM
Didn't see a 2019 Dork Poll thread so feel free to start one.

Anyways....KenPom has us at #2 and the ACC heavily repped in the top 25.

And what's up with Nevada?

https://kenpom.com

I don't ask for much, Santa, but please keep our KenPom ranking above UNC, Villanova, and Virginia throughout the year. And, while you're at it, keep Kentucky below Nevada.

Wahoo2000
10-22-2018, 01:32 PM
I don't ask for much, Santa, but please keep our KenPom ranking above UNC, Villanova, and Virginia throughout the year. And, while you're at it, keep Kentucky below Nevada.

If you beat UK to start the season, you'll go a long way towards getting your wishes (at least through Christmas and 2018).

kAzE
10-22-2018, 01:37 PM
Not impossible that some mid-majors could send more teams to the tourney than the "conference of champions".

*NIT champions

flyingdutchdevil
10-22-2018, 01:41 PM
I don't ask for much, Santa, but please keep our KenPom ranking above UNC, Villanova, and Virginia throughout the year. And, while you're at it, keep Kentucky below Nevada.

I wouldn't put your money on us being above Virginia...

TexHawk
10-22-2018, 03:03 PM
Kansas' projection is somewhat interesting. While many places have them #1, here kenpom is projecting dominance from the Jawhawks. #2 offensive team, #1 defensive team. The difference between their projected efficiency margin and the #2 team Duke's is approximately the difference between Duke and the #10 team WVU, or the difference between WVU and the #26 team Texas Tech.

I think Kenpom is factoring transfers into his preseason formula for the first time this year, and 60% of KU's likely starting 5 are transfers. (And one of the others is a freshman.) I have a feeling that this projection is going to be (at least) mildly off for that reason.

Dedric Lawson is going to be very good, but I'd be surprised if he puts up the same stats at KU that he did at Memphis. That would be Bagley/Ayton-like numbers, which would be a little aggressive imo.

HereBeforeCoachK
10-22-2018, 03:34 PM
I wouldn't put your money on us being above Virginia...

...perhaps, but I would put money on us playing longer than Virginia....

House P
10-29-2018, 09:55 AM
Based on individual game predictions, KenPom estimates that there is about a 40% chance that Duke's current non-conference home winning streak will end this season.

This is based primarily on the fact that Duke plays home games against two teams ranked in KenPom's preseason top 40 (Indiana and St. Johns) and three home games against teams ranked between 100 and 150 (Eastern Michigan, Yale, and Princeton).

Here is how KenPom currently estimates Duke odds of winning each home game.



Opponent
KenPom
Rank
Predicted
Point Spread
Chance of
Duke Winning


Indiana
29
Duke -9
81%


St. Johns
39
Duke -12
86%


E Michigan
113
Duke -18
95%


Yale
116
Duke -19
96%


Princeton
148
Duke -21
97%


Hartford
197
Duke -22
99%


Army
223
Duke -26
99%


Stetson
334
Duke -34
99.9%





















The obvious caveat is that ratings pre-season ratings should probably be taken with a grain of salt.

For example, consider last season's games vs Elon and NC State. Based on KenPom's preseason rankings, both teams appeared to be of similar strength (Elon #106, NC State #109). However, Elon went on to have a disappointing 14-18 season and ended the year ranked #247 by KenPom. NC State went on to have a solid year, making the NCAA tournament with a final KenPom ranking of #42.

Wahoo2000
11-05-2018, 02:37 PM
Kenpom updated today to reflect late additions via RS waivers, etc. Addition of Braxton Key pushed UVa just a fingernail above Duke and UNC and into the #2 spot.

What's REALLY interesting is that the 3 teams are within .18 Adjusted Efficiency Margin points of each other... meaning that a neutral floor contest among any of them is basically a pickem, or a 1pt line at most. With quality PROJECTING to be so equal, this could be the FIRST time (despite TONS of yelling and screaming over the past few seasons.... I don't want to get into the data and metrics of it all, you just have to "trust me" ;) ) that conference SOS really COULD influence who becomes the regular season champion.

flyingdutchdevil
11-05-2018, 02:47 PM
Kenpom updated today to reflect late additions via RS waivers, etc. Addition of Braxton Key pushed UVa just a fingernail above Duke and UNC and into the #2 spot.

What's REALLY interesting is that the 3 teams are within .18 Adjusted Efficiency Margin points of each other... meaning that a neutral floor contest among any of them is basically a pickem, or a 1pt line at most. With quality PROJECTING to be so equal, this could be the FIRST time (despite TONS of yelling and screaming over the past few seasons... I don't want to get into the data and metrics of it all, you just have to "trust me" ;) ) that conference SOS really COULD influence who becomes the regular season champion.

You mean between UNC and UVA? Duke is the most talented, but they will have plenty of stumbles in the ACC. Duke may have a chance to win, but it's a small chance.

UVA is winning the reg ACC. I can't really see any scenario where that doesn't happen. UNC is good, and Little is the real deal, but I don't see them competing with the uber disciplined team that is UVA.

Troublemaker
11-05-2018, 02:57 PM
You mean between UNC and UVA? Duke is the most talented, but they will have plenty of stumbles in the ACC. Duke may have a chance to win, but it's a small chance.

UVA is winning the reg ACC. I can't really see any scenario where that doesn't happen. UNC is good, and Little is the real deal, but I don't see them competing with the uber disciplined team that is UVA.

You take UVA to be the 1 seed in the ACC tourney, and I'll take the field. Bet a pie?

Wahoo2000
11-05-2018, 03:00 PM
You mean between UNC and UVA? Duke is the most talented, but they will have plenty of stumbles in the ACC. Duke may have a chance to win, but it's a small chance.

UVA is winning the reg ACC. I can't really see any scenario where that doesn't happen. UNC is good, and Little is the real deal, but I don't see them competing with the uber disciplined team that is UVA.

I think you might be slightly influenced by your heavily frosh laden teams from the past 4-5 seasons. This team is different. The typical chasm in talent between a top 3 prospect and a top 10-20 prospect is huuuuuuuuuuuuge. The former are guys that become NBA all-stars. These guys can dominate in college from day 1. Now, if you have just ONE of those players, and surround them with top 5-25 guys, you're probably not talented enough to overcome some "stumbles" w/r/t continuity, team chemistry, learning defense, etc. If you have 3 of those guys? The margin for error becomes OH SO MUCH greater. I'm sure many here feel differently after first hand witnessing the "struggles" of past Duke "uber talented" teams. But I'm telling you right now, that barring injury issues (pre-bolding that so someone doesn't have to bother when they inevitably respond with quotes and say, "that's it right there"), this Duke team is on another level from a talent standpoint. An "omega-talented" team, if I can coin a phrase.

All I'm saying is - this team could have the EXACT same issues with learning defense, chemistry, etc as the previous ones. They WILL NOT be losing head-scratchers to Miami/NC State/BC/St Johns/etc. You guys might drop one to KY tomorrow, as I think they're a legit top 5 team, and those can go either way. Don't let that temper your enthusiasm or expectations. I'm gonna go out on a limb and say Duke will have absolutely ZERO "wtf" losses this year. Just waaaaaaaaaaay too much talent. Enough that the 14-18 Duke teams would go, "DAMN! That's a lot of talent!"


-Wahoo"On rereading I realized that I might have set the record for unnecessary overuse of quotation marks in one post"2000

kAzE
11-05-2018, 03:41 PM
All I'm saying is - this team could have the EXACT same issues with learning defense, chemistry, etc as the previous ones. They WILL NOT be losing head-scratchers to Miami/NC State/BC/St Johns/etc. You guys might drop one to KY tomorrow, as I think they're a legit top 5 team, and those can go either way. Don't let that temper your enthusiasm or expectations. I'm gonna go out on a limb and say Duke will have absolutely ZERO "wtf" losses this year. Just waaaaaaaaaaay too much talent. Enough that the 14-18 Duke teams would go, "DAMN! That's a lot of talent!"


-Wahoo"On rereading I realized that I might have set the record for unnecessary overuse of quotation marks in one post"2000

I'd bet against it. We will definitely have at least 1 or 2 really dumb losses this year. Maybe not to Yale or Princeton at home, but it always happens, when you least expect it. But it won't be something to stress over. We've basically conceded all regular season titles with the OAD recruiting strategy. Even the 2014-15 team had a crazy bad WTF loss on the road against a garbage NC State team.

I'm looking hard at that first true road game @ Wake Forest. I think that would absolutely qualify as a "WTF" loss.

budwom
11-05-2018, 03:58 PM
I'd bet against it. We will definitely have at least 1 or 2 really dumb losses this year. Maybe not to Yale or Princeton at home, but it always happens, when you least expect it. But it won't be something to stress over. We've basically conceded all regular season titles with the OAD recruiting strategy. Even the 2014-15 team had a crazy bad WTF loss on the road against a garbage NC State team.

I'm looking hard at that first true road game @ Wake Forest. I think that would absolutely qualify as a "WTF" loss.

I agree 100%, Kaze...could be an epic season (or not), but there will most assuredly be ugly bumps in the road....not easy hitting the road in the ACC with a gang of youngsters, their fine talent notwithstanding.
I don't even know our schedule all that well nor our opponents; rosters, but it rarely seems easy going to places like Tallahassee, Raleigh, Clemmons et al...

flyingdutchdevil
11-05-2018, 04:21 PM
I think you might be slightly influenced by your heavily frosh laden teams from the past 4-5 seasons. This team is different. The typical chasm in talent between a top 3 prospect and a top 10-20 prospect is huuuuuuuuuuuuge. The former are guys that become NBA all-stars. These guys can dominate in college from day 1. Now, if you have just ONE of those players, and surround them with top 5-25 guys, you're probably not talented enough to overcome some "stumbles" w/r/t continuity, team chemistry, learning defense, etc. If you have 3 of those guys? The margin for error becomes OH SO MUCH greater. I'm sure many here feel differently after first hand witnessing the "struggles" of past Duke "uber talented" teams. But I'm telling you right now, that barring injury issues (pre-bolding that so someone doesn't have to bother when they inevitably respond with quotes and say, "that's it right there"), this Duke team is on another level from a talent standpoint. An "omega-talented" team, if I can coin a phrase.

All I'm saying is - this team could have the EXACT same issues with learning defense, chemistry, etc as the previous ones. They WILL NOT be losing head-scratchers to Miami/NC State/BC/St Johns/etc. You guys might drop one to KY tomorrow, as I think they're a legit top 5 team, and those can go either way. Don't let that temper your enthusiasm or expectations. I'm gonna go out on a limb and say Duke will have absolutely ZERO "wtf" losses this year. Just waaaaaaaaaaay too much talent. Enough that the 14-18 Duke teams would go, "DAMN! That's a lot of talent!"


-Wahoo"On rereading I realized that I might have set the record for unnecessary overuse of quotation marks in one post"2000

I think you're underestimating the impact of youth. The 2015 team - stacked with talent and some amazing leadership in Cook and Jefferson - got murdered by Miami at home. And that team had the best big man in the country, on of the most productive PGs in the country, a 15ppg senior leader, and a Swiss-army knife in Winslow.

This year, there will be a "WTF" loss (and please patent this phrase. It's amazing).

Every year, we tell ourselves this year is unstoppable. In 2016-17, we had the top 2 players in the country (and this time 2 years ago, many were saying Giles would be 100% by February/March) coupled with Grayson, Jefferson, and countless OADs. In 2017-18, we thought we had the most perfect starting 5 with the top player at 4 positions and Grayson as our senior leader who can score 20 a game. This year, we are staying that ESPN's 1-3 top players means we have too much talent. But we lack depth and we lack experience. And those will lead to at least one "WTF" loss.

BandAlum83
11-05-2018, 04:25 PM
You mean between UNC and UVA? Duke is the most talented, but they will have plenty of stumbles in the ACC. Duke may have a chance to win, but it's a small chance.

UVA is winning the reg ACC. I can't really see any scenario where that doesn't happen. UNC is good, and Little is the real deal, but I don't see them competing with the uber disciplined team that is UVA.

I really don't care. I just want to be sure we get the double bye!

BandAlum83
11-05-2018, 04:29 PM
I think you might be slightly influenced by your heavily frosh laden teams from the past 4-5 seasons. This team is different. The typical chasm in talent between a top 3 prospect and a top 10-20 prospect is huuuuuuuuuuuuge. The former are guys that become NBA all-stars. These guys can dominate in college from day 1. Now, if you have just ONE of those players, and surround them with top 5-25 guys, you're probably not talented enough to overcome some "stumbles" w/r/t continuity, team chemistry, learning defense, etc. If you have 3 of those guys? The margin for error becomes OH SO MUCH greater. I'm sure many here feel differently after first hand witnessing the "struggles" of past Duke "uber talented" teams. But I'm telling you right now, that barring injury issues (pre-bolding that so someone doesn't have to bother when they inevitably respond with quotes and say, "that's it right there"), this Duke team is on another level from a talent standpoint. An "omega-talented" team, if I can coin a phrase.

All I'm saying is - this team could have the EXACT same issues with learning defense, chemistry, etc as the previous ones. They WILL NOT be losing head-scratchers to Miami/NC State/BC/St Johns/etc. You guys might drop one to KY tomorrow, as I think they're a legit top 5 team, and those can go either way. Don't let that temper your enthusiasm or expectations. I'm gonna go out on a limb and say Duke will have absolutely ZERO "wtf" losses this year. Just waaaaaaaaaaay too much talent. Enough that the 14-18 Duke teams would go, "DAMN! That's a lot of talent!"


-Wahoo"On rereading I realized that I might have set the record for unnecessary overuse of quotation marks in one post"2000

I am completely with you on this. We are definitely tempering expectations.

While we do that, we are all secretly thinking about that mythical 40-0 season, because really.....the potential top 3 picks in the next NBA draft, plus a legit pass-first, quality on-ball defending PG? Are you freaking kidding me?

Troublemaker
11-05-2018, 05:12 PM
Even great veteran teams will suffer a WTF loss. Heck, UVA's only loss last season was at home to VaTech, after having beaten them in Blacksburg by 26 the month prior.

So, Wahoo2000 is technically wrong just because almost any team will suffer a WTF loss or two. But I am with Wahoo2000 in spirit in that I think this Duke team will be more consistent than some other recent versions. I think this team's talent fits how Coach K wants to play basketball like a glove, and I believe in the power of the steady, true PG.

kAzE
11-05-2018, 05:20 PM
Even great veteran teams will suffer a WTF loss. Heck, UVA's only loss last season was at home to VaTech, after having beaten them in Blacksburg by 26 the month prior.

So, Wahoo2000 is technically wrong just because almost any team will suffer a WTF loss or two. But I am with Wahoo2000 in spirit in that I think this Duke team will be more consistent than some other recent versions. I think this team's talent fits how Coach K wants to play basketball like a glove, and I believe in the power of the steady, true PG.

UMBC was pretty "WTF"

Sorry, I had to go there :cool:

Troublemaker
11-05-2018, 05:20 PM
UVA is winning the reg ACC. I can't really see any scenario where that doesn't happen.


You take UVA to be the 1 seed in the ACC tourney, and I'll take the field. Bet a pie?

FDD, you want that pie bet or not? If, instead, you want to say that you were a bit hasty in not being able to envision "any scenario," nobody will think any less of you.

NSDukeFan
11-05-2018, 07:58 PM
I think you're underestimating the impact of youth. The 2015 team - stacked with talent and some amazing leadership in Cook and Jefferson - got murdered by Miami at home. And that team had the best big man in the country, on of the most productive PGs in the country, a 15ppg senior leader, and a Swiss-army knife in Winslow.

This year, there will be a "WTF" loss (and please patent this phrase. It's amazing).

Every year, we tell ourselves this year is unstoppable. In 2016-17, we had the top 2 players in the country (and this time 2 years ago, many were saying Giles would be 100% by February/March) coupled with Grayson, Jefferson, and countless OADs. In 2017-18, we thought we had the most perfect starting 5 with the top player at 4 positions and Grayson as our senior leader who can score 20 a game. This year, we are staying that ESPN's 1-3 top players means we have too much talent. But we lack depth and we lack experience. And those will lead to at least one "WTF" loss.

I only thought we were unstoppable in 2010-2011 and 2016-17, when we had great talent and experience. Injuries, grrrr!

Wahoo2000
11-07-2018, 12:12 AM
I think you might be slightly influenced by your heavily frosh laden teams from the past 4-5 seasons. This team is different. The typical chasm in talent between a top 3 prospect and a top 10-20 prospect is huuuuuuuuuuuuge. The former are guys that become NBA all-stars. These guys can dominate in college from day 1. Now, if you have just ONE of those players, and surround them with top 5-25 guys, you're probably not talented enough to overcome some "stumbles" w/r/t continuity, team chemistry, learning defense, etc. If you have 3 of those guys? The margin for error becomes OH SO MUCH greater. I'm sure many here feel differently after first hand witnessing the "struggles" of past Duke "uber talented" teams. But I'm telling you right now, that barring injury issues (pre-bolding that so someone doesn't have to bother when they inevitably respond with quotes and say, "that's it right there"), this Duke team is on another level from a talent standpoint. An "omega-talented" team, if I can coin a phrase.

All I'm saying is - this team could have the EXACT same issues with learning defense, chemistry, etc as the previous ones. They WILL NOT be losing head-scratchers to Miami/NC State/BC/St Johns/etc. You guys might drop one to KY tomorrow, as I think they're a legit top 5 team, and those can go either way. Don't let that temper your enthusiasm or expectations. I'm gonna go out on a limb and say Duke will have absolutely ZERO "wtf" losses this year. Just waaaaaaaaaaay too much talent. Enough that the 14-18 Duke teams would go, "DAMN! That's a lot of talent!"


-Wahoo"On rereading I realized that I might have set the record for unnecessary overuse of quotation marks in one post"2000

I'm just going to leave this riiiiight here after watching this opener.

Again, you CANNOT compare this Duke team to any of the other recent ones. This is like having THREE Tatums, or THREE Ingrams, or THREE Bagleys. Not those elite guys surrounded by other "pretty good" (and by pretty good, I mean still potential lottery pick players) guys.

I WILL say that I look forward to seeing our matchup again.
Guy on Jones
Jerome on Reddish
Key on Barrett
Hunter on Zion
Salt v Bolden
I don't think any team in the nation will hold Duke under 70, but if there is one, we're probably it. Until then - good health and good luck (though it doesn't look like you'll need the luck!)

DavidBenAkiva
12-10-2018, 08:19 AM
Something interesting is happening on KenPom and the other advanced analytics ranking systems, AKA dork polls. Duke is pulling ahead of the field.

The trend started before the weekend. As of this morning (12/10), Duke's adjusted efficiency margin on KenPom is far ahead of the second place team, Kansas. In fact, the difference between Duke and Kansas is about the same as the difference between Kansas and the 11th-placed team, Virginia Tech. A lot of this is driven by Duke's stellar defensive effort this season. The combination of blocked shots, steals, and shutting down the shooting percentages of opponents has Duke with both a top 5 offense and defense in the country.

It's a similar story on Bart Torvik, who runs a free version of what KenPom does. There, Torvik has Duke at #1 with Michigan at #2 by a healthy gap. Team Rankings (https://www.teamrankings.com/ncaa-basketball/ranking/predictive-by-other/) also runs a predictive rankings model and has Duke at #1, as does Sagarin. After Gonzaga lost, Duke also took the #1 spot on ESPN's BPI. So the stats are converging that this is a special team. I sure hope that Duke can keep it up.

MChambers
12-10-2018, 08:29 AM
I'm just going to leave this riiiiight here after watching this opener.

Again, you CANNOT compare this Duke team to any of the other recent ones. This is like having THREE Tatums, or THREE Ingrams, or THREE Bagleys. Not those elite guys surrounded by other "pretty good" (and by pretty good, I mean still potential lottery pick players) guys.

I WILL say that I look forward to seeing our matchup again.
Guy on Jones
Jerome on Reddish
Key on Barrett
Hunter on Zion
Salt v Bolden
I don't think any team in the nation will hold Duke under 70, but if there is one, we're probably it. Until then - good health and good luck (though it doesn't look like you'll need the luck!)

What about Clark? Isn't he playing much more than Key? So you would put Clark on Jones, I assume, and then one of Guy or Jerome has to be on Barrett, right?

Or do you think Clark just gets far fewer minutes against Duke?

freshmanjs
12-10-2018, 08:44 AM
A lot of this is driven by Duke's stellar defensive effort this season. The combination of blocked shots, steals, and shutting down the shooting percentages of opponents has Duke with both a top 5 offense and defense in the country.

In Kenpom, the gap between Duke and Kansas is driven more by offense than defense. #3 Virginia and #4 Michigan have better ranked defense than Duke.

Our separation vs. Gonzaga and Nevada is driven by defense right now.

Of course, it's the combination that has us at the top.

sagegrouse
12-10-2018, 08:49 AM
You mean between UNC and UVA? Duke is the most talented, but they will have plenty of stumbles in the ACC. Duke may have a chance to win, but it's a small chance.

UVA is winning the reg ACC. I can't really see any scenario where that doesn't happen. UNC is good, and Little is the real deal, but I don't see them competing with the uber disciplined team that is UVA.

Oh my! Literally untrue. "Any scenario?" How about Duke wins at Cameron and in Charlottesville? Not possible? I bet there are UVa fans that think it is highly possible.

HereBeforeCoachK
12-10-2018, 09:55 AM
Oh my! Literally untrue. "Any scenario?" How about Duke wins at Cameron and in Charlottesville? Not possible? I bet there are UVa fans that think it is highly possible.

I wonder if that poster was one of several who was saying that UK was going to beat us by 15 or so before the season opener.......

Wahoo2000
12-10-2018, 11:01 AM
What about Clark? Isn't he playing much more than Key? So you would put Clark on Jones, I assume, and then one of Guy or Jerome has to be on Barrett, right?

Or do you think Clark just gets far fewer minutes against Duke?

A lot has happened since my post from early Nov.

1 - I don't think UVA is playing anywhere nearly as defensively sound as I'd expected with so many returnees. Could be partly because we've added some wrinkles (actually running some press). Haven't seen a UVA team this "shaky" on D (if shaky means more like a top 5-10ish D rather than clear cut #1 or #2) since the 1st half of Brogdon's last year. We didn't really start to lock in defensively until around the second pass through the ACC schedule. Also similar to that team, our offense has been much better than in most seasons - in fact, if we shot even a reasonable percentage on the 6-7 wide open 3s we got vs VCU yesterday, we're probably even better.
2 - Clark has really emerged as a major rotation player - probably 4th in the playing time "pecking order" after Jerome/Guy/Hunter. He is an absolute bulldog defensively. I've yet to see him really get beaten off the bounce, and even forced a 10 second violation with absolutely NO help, including having to get around a screen in the backcourt. He IS set to have wrist surgery today, and it's an unknown how long he'll be out. I'd assume at least 3-4 weeks and prob return with a cast (like the one he played with yesterday) for a few weeks after that.
3 - Not sure how we'll match up with Duke if both Clark and Guy play major minutes. Could be a field day for whichever of Barrett/Reddish draw Guy as a defender. Guy is solid fundamentally, but not the quickest, biggest, or strongest. Honestly, if it doesn't "click" a little better for Key and Diakite by then, I think we're in BIG trouble unless we can really just pack the paint and hope Duke shoots like 25% from 3.

Troublemaker
12-10-2018, 11:16 AM
What about Clark? Isn't he playing much more than Key? So you would put Clark on Jones, I assume, and then one of Guy or Jerome has to be on Barrett, right?

Or do you think Clark just gets far fewer minutes against Duke?


Oh my! Literally untrue. "Any scenario?" How about Duke wins at Cameron and in Charlottesville? Not possible? I bet there are UVa fans that think it is highly possible.

It should be noted that some of these posts were made more than a month ago.

Hopefully FDD's prediction doesn't age well.


Something interesting is happening on KenPom and the other advanced analytics ranking systems, AKA dork polls. Duke is pulling ahead of the field.

The trend started before the weekend. As of this morning (12/10), Duke's adjusted efficiency margin on KenPom is far ahead of the second place team, Kansas. In fact, the difference between Duke and Kansas is about the same as the difference between Kansas and the 11th-placed team, Virginia Tech. A lot of this is driven by Duke's stellar defensive effort this season. The combination of blocked shots, steals, and shutting down the shooting percentages of opponents has Duke with both a top 5 offense and defense in the country.

It's a similar story on Bart Torvik, who runs a free version of what KenPom does. There, Torvik has Duke at #1 with Michigan at #2 by a healthy gap. Team Rankings (https://www.teamrankings.com/ncaa-basketball/ranking/predictive-by-other/) also runs a predictive rankings model and has Duke at #1, as does Sagarin. After Gonzaga lost, Duke also took the #1 spot on ESPN's BPI. So the stats are converging that this is a special team. I sure hope that Duke can keep it up.

Thanks for pointing this out, DBA. I guess the question then becomes: do we believe it? Do we believe that Duke belongs in a tier of its own at the top of college basketball? I'm open to having my mind changed down the line, but as of now, my answer is "No." Certainly I believe we're a top-tier team but I believe there are other occupants of that tier including Michigan, UVA, maybe Kansas when healthy, maybe Nevada, maybe Gonzaga when Tillie returns. I think what may be happening statistically is that Duke is better at blowing teams out than other top-tier teams because of our pressure defense and devastating transition game, but if Michigan and Duke played a 7-game series, I would consider that to be a coin-flip series at this point, not advantage Duke.

HereBeforeCoachK
12-10-2018, 11:18 AM
I guess the question then becomes: do we believe it? Do we believe that Duke belongs in a tier of its own at the top of college basketball? I'm open to having my mind changed down the line, but as of now, my answer is "No." Certainly I believe we're a top-tier team but I believe there are other occupants of that tier including Michigan, UVA, maybe Kansas when healthy, maybe Nevada, maybe Gonzaga when Tillie returns. I think what may be happening statistically is that Duke is better at blowing teams out than other top-tier teams because of our pressure defense and devastating transition game, but if Michigan and Duke played a 7-game series, I would consider that to be a coin-flip series at this point, not advantage Duke.

You're on a roll....another great post! (I'll try and spork). Great point about being better at blowing people out....which will help Kenpom, but NET, not so much. I'm thinking I'm a little more willing to call Duke top tier than you are at this point, but your points are well written and well taken.

Troublemaker
12-10-2018, 11:26 AM
3 - Not sure how we'll match up with Duke if both Clark and Guy play major minutes. Could be a field day for whichever of Barrett/Reddish draw Guy as a defender. Guy is solid fundamentally, but not the quickest, biggest, or strongest. Honestly, if it doesn't "click" a little better for Key and Diakite by then, I think we're in BIG trouble unless we can really just pack the paint and hope Duke shoots like 25% from 3.

Yeah, there's no way you guys play that small when it's time to play Duke. With Hunter and Key, you can match up defensively with Duke's big NBA wings better than the almost any school in the country. It'd be mismanagement not to give Key starter's minutes against Duke, which I don't expect Bennett will be guilty of. Can you imagine Clark or Guy trying to keep Barrett off the boards? (RJ's rebounding prowess is perhaps the most underrated skill on Duke's team right now).

DavidBenAkiva
12-10-2018, 11:33 AM
You're on a roll...another great post! (I'll try and spork). Great point about being better at blowing people out...which will help Kenpom, but NET, not so much. I'm thinking I'm a little more willing to call Duke top tier than you are at this point, but your points are well written and well taken.

Ditto. As far as we are willing to trust the stats, Duke is a very good team. Even with the poor shooting of late, we are blowing out teams by 20+. More impressively, we are absolutely shutting down opposing NBA-level prospects. Romeo Langford was completely ineffective against Duke. Miye Oni of Yale is a legit NBA prospect and he was held to 12 points and looked completely outmatched against R.J. Barrett and crew. That will be a test for the offense as the Red Raiders have been playing defense at a high level. They are among the leaders in the nation in effective field goal defense. I hope Duke can add another signature neutral court victory to their resume.

JasonEvans
12-10-2018, 11:49 AM
Thanks for pointing this out, DBA. I guess the question then becomes: do we believe it? Do we believe that Duke belongs in a tier of its own at the top of college basketball?

It is worth noting that the "Duke is on a different level" thing is not just happening on KenPom. In Sagarin's rankings, Duke is a massive 4.6 points ahead of 2nd place Michigan State. That's the same as the gap between MSU and the #13 team in the Sagarin rankings.

Worth noting that ESPN's BPI is not showing the same kind of "Duke and everyone else" gap. Duke is #1, but only by 0.3 over#2 UVA. The gap between UVA and #3 UNC is 0.7.

I am inclined to agree with my Troublemaking friend that it is too soon to say that Duke is on a different level from the rest of the sport. There simply have not been enough games between top tier teams to know that. But, once Duke plays Texas Tech (12/20), and then runs that @FSU (1/12), Syr (1/14), UVA (1/18) mini-gauntlet we will probably have enough evidence to really know if this is a prohibitive NCAA favorite or just one of several teams that seem like they have a decent shot at winning the title.

-Jason "strap in folks... it could be a fun ride!" Evans

Ian
12-10-2018, 12:20 PM
Under Coach K, there have only been 3 teams where you could say, "if they don't win the title it's a major disappointment", thoses teams are 1992, 1999, and 2002.
Three other teams come close to fitting that description. 2001, 2011, and 1986, but don't quite make it. 2001 and 2011 would have had those expectations healthy but because of the injuries to Boozer and Irving their expectations got somewhat diminished, and 1986 because it was Coach K's first great team fans were more "just happy to be here" and not disappointed with anything.

I think there is a good chance that this team is the 4th team to have that kind of pressure on them.

English
12-10-2018, 12:21 PM
...Thanks for pointing this out, DBA. I guess the question then becomes: do we believe it? Do we believe that Duke belongs in a tier of its own at the top of college basketball? I'm open to having my mind changed down the line, but as of now, my answer is "No." Certainly I believe we're a top-tier team but I believe there are other occupants of that tier including Michigan, UVA, maybe Kansas when healthy, maybe Nevada, maybe Gonzaga when Tillie returns. I think what may be happening statistically is that Duke is better at blowing teams out than other top-tier teams because of our pressure defense and devastating transition game, but if Michigan and Duke played a 7-game series, I would consider that to be a coin-flip series at this point, not advantage Duke.

I agree with most of your larger point, but I'd actually feel much better about Duke's chances in a 7-game series, as opposed to a one-game matchup (even on a neutral court). Beyond the coaching advantage I'd see our guys having, I think the Duke team is much more poised to exploit advantages and avoid potential hot streaks in a series. For example, there's no way Gonzaga shoots 60% from the field and 60% from 3-pt in a 7-game series, and yet here we are with a loss to them in a one-game sample. Even with Tilley back, I'd be more confident over seven games than the prospect of someone like Norvell catching fire in one game. I'd say the same of Michigan, UVa, and really, anyone else.

Of course, the closest we may ever get to this would be a 4-game series with UVa (or the morally bankrupt team down the road), and even then, it wouldn't be quite the same setup.

HereBeforeCoachK
12-10-2018, 12:21 PM
Ditto. As far as we are willing to trust the stats, Duke is a very good team. Even with the poor shooting of late, we are blowing out teams by 20+. More impressively, we are absolutely shutting down opposing NBA-level prospects. Romeo Langford was completely ineffective against Duke. Miye Oni of Yale is a legit NBA prospect and he was held to 12 points and looked completely outmatched against R.J. Barrett and crew. That will be a test for the offense as the Red Raiders have been playing defense at a high level. They are among the leaders in the nation in effective field goal defense. I hope Duke can add another signature neutral court victory to their resume.

Texas Tech is real. They coulda/shoulda beaten Nova in the Elite 8 last season....as this was Villanova's worst game IMO.....I really think TT didn't realize they were good enough to win, but they were.

BandAlum83
12-10-2018, 12:35 PM
Under Coach K, there have only been 3 teams where you could say, "if they don't win the title it's a major disappointment", thoses teams are 1992, 1999, and 2002.
Three other teams come close to fitting that description. 2001, 2011, and 1986, but don't quite make it. 2001 and 2011 would have had those expectations healthy but because of the injuries to Boozer and Irving their expectations got somewhat diminished, and 1986 because it was Coach K's first great team fans were more "just happy to be here" and not disappointed with anything.

I think there is a good chance that this team is the 4th team to have that kind of pressure on them.

The loss in the final is, to this day, still my most painful duke sports memory! We were absolutely the best team! Maybe things evened out with us winning in 2010 without the best team, but '86 still hurts!

HereBeforeCoachK
12-10-2018, 12:38 PM
, and 1986 because it was Coach K's first great team fans were more "just happy to be here" and not disappointed with anything.
.

Au contraire sir, with due respect. That team had been building for years, from good, to very good, to great throughout 86. Clearly the best team in the country (and Kansas was clearly second best). I agree with some other posters who rate this as the most painful loss.....not only because they were clearly the best, but what a great group of people - and this goes back before Duke was hated universally too....

JasonEvans
12-10-2018, 01:04 PM
I was a freshman in 1986 and 30+ years later that loss is still the most painful I have endured as a Duke fan.

-Jason "the 1999 championship game comes in a close second as most gut-wrenching losses... both times we were far and away the best team in the land. Sigh" Evans

uh_no
12-10-2018, 01:43 PM
I was a freshman in 1986 and 30+ years later that loss is still the most painful I have endured as a Duke fan.

-Jason "the 1999 championship game comes in a close second as most gut-wrenching losses... both times we were far and away the best team in the land. Sigh" Evans

While KP doesn't go back that far, I'd image duke would have been #1 by a large margin, and uconn probably would have been #2. The degree to which duke regularly beat down good teams down the stretch was incredible. That said, in hindsight, i might give duke 2 out of 3 over connecticut. 65% sounds about right. maybe 3/5.

DavidBenAkiva
12-10-2018, 01:44 PM
Texas Tech is real. They coulda/shoulda beaten Nova in the Elite 8 last season...as this was Villanova's worst game IMO....I really think TT didn't realize they were good enough to win, but they were.

Just a quick note on Texas Tech: The team has had a lot of roster turnover since last season, as much as Duke. It's practically a new team. Freshman Zhaire Smith was a 1st round NBA Draft pick. He was among 6 of their 8 leading scorers that left the team after the season. The other 7 were all seniors. The team this year was unranked at the start of the season as a result. Texas Tech added a number of grad transfers, including center Tariq Owens, who played at St. John's last season (and was particularly effective in the game against Duke) and guard Matt Mooney from South Dakota.

Reilly
12-10-2018, 01:45 PM
... 1986 ... 1999 championship game comes in a close second ... both times we were far and away the best team in the land ...

I imagine many will agree with you about the 1999 team being far and away the best team in the land, but many would say 1986 was closer to the pack.

1999: 34.80 SRS ... next closest team was MD at 26.19 (8+ points worse) ... UConn was 24.74 (10+ points worse)

1986: 22.10 SRS ... Duke was #3 in the country per this computer, behind UNC and Kansas. L'ville was #8 at 20.64.

https://www.sports-reference.com/cbb/schools/duke/

MChambers
12-10-2018, 02:02 PM
Texas Tech is real. They coulda/shoulda beaten Nova in the Elite 8 last season...as this was Villanova's worst game IMO....I really think TT didn't realize they were good enough to win, but they were.

This is a very different Texas Tech team. Maybe just as good, but they lost 3 or 4 starters.

MChambers
12-10-2018, 02:04 PM
A lot has happened since my post from early Nov.

1 - I don't think UVA is playing anywhere nearly as defensively sound as I'd expected with so many returnees. Could be partly because we've added some wrinkles (actually running some press). Haven't seen a UVA team this "shaky" on D (if shaky means more like a top 5-10ish D rather than clear cut #1 or #2) since the 1st half of Brogdon's last year. We didn't really start to lock in defensively until around the second pass through the ACC schedule. Also similar to that team, our offense has been much better than in most seasons - in fact, if we shot even a reasonable percentage on the 6-7 wide open 3s we got vs VCU yesterday, we're probably even better.
2 - Clark has really emerged as a major rotation player - probably 4th in the playing time "pecking order" after Jerome/Guy/Hunter. He is an absolute bulldog defensively. I've yet to see him really get beaten off the bounce, and even forced a 10 second violation with absolutely NO help, including having to get around a screen in the backcourt. He IS set to have wrist surgery today, and it's an unknown how long he'll be out. I'd assume at least 3-4 weeks and prob return with a cast (like the one he played with yesterday) for a few weeks after that.
3 - Not sure how we'll match up with Duke if both Clark and Guy play major minutes. Could be a field day for whichever of Barrett/Reddish draw Guy as a defender. Guy is solid fundamentally, but not the quickest, biggest, or strongest. Honestly, if it doesn't "click" a little better for Key and Diakite by then, I think we're in BIG trouble unless we can really just pack the paint and hope Duke shoots like 25% from 3.

Sorry, read your post as being 12/7, not 11/7. I am surprised by how much time Clark is getting, just because I don't think UVa can go that small against some of the better teams in the country.

JasonEvans
12-10-2018, 02:27 PM
While KP doesn't go back that far, I'd image duke would have been #1 by a large margin, and uconn probably would have been #2. The degree to which duke regularly beat down good teams down the stretch was incredible. That said, in hindsight, i might give duke 2 out of 3 over connecticut. 65% sounds about right. maybe 3/5.

Just to remind folks about the absolute monster that was the 1999 Blue Devils.

@ Ga Tech
@ St John

Those are the only two games from Jan 1 until the Final Four that 1999 Duke played that were not double digit victories. Duke actually went to OT against the Johnnies, winning by 4, and we only beat Ga Tech by 8.

Aside from those two, it was pretty much a massacre every time out. Maryland spent almost the entire season in the top 5 that year... we beat them by about 20 points every time we played them. Carolina was a top 10-15 teams all year... we also ripped them by 20 every time we played. We went undefeated in the ACC that year and set a record for victory margin. It was not uncommon for us to beat ACC teams by 30 or 40 points. We were on a 32 game winning streak when we fell to UConn.

As for UConn, they were really good in 1999, but nowhere near the level of dominance we saw from Duke. In Feburary they lost by 17 points (at home) to a decent but not great Syracuse team. They lost to Miami too. They had a nice win over a top 5 Stanford club by 11 points, but had a number of nailbiters against just so-so teams like Seton Hall and Providence.

I don't know what goes into KenPom's probability stats, but I suspect he would have had Duke more like 80% to win that national title game.

Matches
12-10-2018, 02:30 PM
As for UConn, they were really good in 1999, but nowhere near the level of dominance we saw from Duke. In Feburary they lost by 17 points (at home) to a decent but not great Syracuse team. They lost to Miami too. They had a nice win over a top 5 Stanford club by 11 points, but had a number of nailbiters against just so-so teams like Seton Hall and Providence.



IIRC they were shorthanded in both of the games they lost. I don't think they lost a game at full strength all season.

But yeah, they didn't destroy teams the way Duke did. '99 Duke and '91 UNLV are the only teams I've ever seen who just blew people away like that, game after game.

robed deity
12-10-2018, 02:34 PM
Just to remind folks about the absolute monster that was the 1999 Blue Devils.

@ Ga Tech
@ St John

Those are the only two games from Jan 1 until the Final Four that 1999 Duke played that were not double digit victories. Duke actually went to OT against the Johnnies, winning by 4, and we only beat Ga Tech by 8.

Aside from those two, it was pretty much a massacre every time out. Maryland spent almost the entire season in the top 5 that year... we beat them by about 20 points every time we played them. Carolina was a top 10-15 teams all year... we also ripped them by 20 every time we played. We went undefeated in the ACC that year and set a record for victory margin. It was not uncommon for us to beat ACC teams by 30 or 40 points. We were on a 32 game winning streak when we fell to UConn.

As for UConn, they were really good in 1999, but nowhere near the level of dominance we saw from Duke. In Feburary they lost by 17 points (at home) to a decent but not great Syracuse team. They lost to Miami too. They had a nice win over a top 5 Stanford club by 11 points, but had a number of nailbiters against just so-so teams like Seton Hall and Providence.

I don't know what goes into KenPom's probability stats, but I suspect he would have had Duke more like 80% to win that national title game.

My most painful Duke loss, as I was only 5 in '86. 2004 is a distant second, so yeah, I dont like UCONN.

What I remember from '99 though, is that something seemed slightly off about the team during the Southwest Missouri State game and going forward. I saw the Tulsa game (2nd Rd) in person, and that was the best I've ever seen a college team play.

Ian
12-10-2018, 02:42 PM
I was a freshman in 1986 and 30+ years later that loss is still the most painful I have endured as a Duke fan.

-Jason "the 1999 championship game comes in a close second as most gut-wrenching losses... both times we were far and away the best team in the land. Sigh" Evans

This isn't just a reply to you but to the general sentiment in several posts including yours. But I wasn't really focused on how "painful" a loss was as the evalutor. My personal most painful losses were 1999 and 2004, and 2004 didn't make the list at all. It was more of how the team is perceived by both the Duke fanbase and the college basketball universe in general both during the season and particularly going into March. In order to fit the criteria of "not winning the title being a major disappointment", a team has to be the clear favorite to win it all, and maintained that status during most of the regular season.

I really think by that criteria, only 92, 99, and 02 fits. 01, 11, and 86 comes close but doesn't quite get there. I believe there is a good chance the 19 team will be another one of those teams.

uh_no
12-10-2018, 03:00 PM
I don't know what goes into KenPom's probability stats, but I suspect he would have had Duke more like 80% to win that national title game.

That's a huge stretch. For a so-far similarly exceptionally rated Duke team, to give you a comparison, duke has only a 72% chance of beating TT next week on a neutral court. I'd venture that uconn team was significantly better than this TT team, even if that duke team was somewhat better than this duke team...so an extra 8% chance of winning (and note the curve is non-linear...to go from 50->58% win is far easier than going from 70->78%) is a bit ridiculous.

That same St. Johns that took Duke to OT...lost to uconn in the big east final by 19. You understate Syracuse, who was in and out of the rankings. Uconn beat them by 22 and 21 points in the two rematches...including at the carrier dome and in MSG.

There's no doubt Duke was likely the better team...but 80% is absurd.

budwom
12-10-2018, 03:28 PM
I didn't like them at all, but that was a very very good UCONN team, solid everywhere. Historically underrated by many Duke fans. And while I fully understand that AP polls probably aren't as good an indicator as KenPom, the Huskies did spend more time as the AP number one team that year than did Duke...(something like 10-8 in weekly polls).

JayZee
12-10-2018, 04:04 PM
I was a freshman in 1986 and 30+ years later that loss is still the most painful I have endured as a Duke fan.

-Jason "the 1999 championship game comes in a close second as most gut-wrenching losses... both times we were far and away the best team in the land. Sigh" Evans

1986 and 1999 in that order in my mind too. For different reasons, clearly.

Those 1986 seniors were so likable and had grown up together. Plus, that defense. And Cameron was particularly amazing those years. Also, that would have been Duke's first NCAA championship in any major sport I think. Though we only had to wait 6 more months for the Duke 1 Akron Zip men's soccer triumph.

1999, we were just so dominant, and it was cathartic to have such an imposing team after those mid 90s years where the wheels got a touch shaky.

For me, I put 1994 next. That would have been so awesome for Grant.

sagegrouse
12-10-2018, 04:17 PM
This isn't just a reply to you but to the general sentiment in several posts including yours. But I wasn't really focused on how "painful" a loss was as the evalutor. My personal most painful losses were 1999 and 2004, and 2004 didn't make the list at all. It was more of how the team is perceived by both the Duke fanbase and the college basketball universe in general both during the season and particularly going into March. In order to fit the criteria of "not winning the title being a major disappointment", a team has to be the clear favorite to win it all, and maintained that status during most of the regular season.

I really think by that criteria, only 92, 99, and 02 fits. 01, 11, and 86 comes close but doesn't quite get there. I believe there is a good chance the 19 team will be another one of those teams.

They all hurt. Let me offer the 2004 semis, where Duke had a seven point lead against UConn with just over two minutes to go in the worst-officiated Final Four game of all time (thus spake Al F. based on inside-the-NCAA views). Against Arkansas in 1994 we had a ten-point lead, but with seven minutes to go. 1999 taught me a lesson. I thought we were a lead-pipe cinch in St. Pete -- never, ever assume anything in the NCAA hoops tournament.

luvdahops
12-10-2018, 05:32 PM
1986 and 1999 in that order in my mind too. For different reasons, clearly.

Those 1986 seniors were so likable and had grown up together. Plus, that defense. And Cameron was particularly amazing those years. Also, that would have been Duke's first NCAA championship in any major sport I think. Though we only had to wait 6 more months for the Duke 1 Akron Zip men's soccer triumph.

1999, we were just so dominant, and it was cathartic to have such an imposing team after those mid 90s years where the wheels got a touch shaky.

For me, I put 1994 next. That would have been so awesome for Grant.

My thinking is very much along the same lines. '85 grad who came back to campus for a few games the next season, and had lived through both the depths of the '82 and '83 campaigns and the beginning of what were arguably Cameron's glory years. The '86 team may not have jumped off the page as dominant, but they won the ACC regular season and tourney in what might have been the deepest and toughest year ever in the conference, and played a challenging non-conference slate as well. Many of their 37 wins were close and hard fought, but there was every reason to believe we were the best team in the country that March/April.

And couldn't agree more on the '99 team. Especially to lose to Calhoun, who absolutely outcoached K that night.

NSDukeFan
12-10-2018, 05:42 PM
The loss in the final is, to this day, still my most painful duke sports memory! We were absolutely the best team! Maybe things evened out with us winning in 2010 without the best team, but '86 still hurts!

I don’t know. That 2010 team was , I believe, a top 5 KenPom team that closed games exceptionally well and was playing very well at the mid to end of the year. I wasn’t predicting they would win for sure, but I definitely thought they had as good a chance as anyone.

The other team that I think had a chance to be a favourite was the 2016-17 team, had they been healthy. They had the best recruiting class, until this year’s, (Giles and Tatum in the same class, with Bolden and Jackson all top 14 recruits, and Javin Deslauriers, another top 40 recruit? Are you kidding me?) That team also had great veteran leadership (Grayson Allen, one of the favourites for NPOY, Amile Jefferson, super solid 5th year senior , who was averaging a double double before getting injured the previous year and already a title winner, Matt Jones, senior and already a title winner). They also had Luke Kennard, coming off a very promising freshman season, Chase Jeter, a top 15 recruit coming back for his sophomore season. Who was going to beat that team, if healthy, physically and emotionally? That was the best I felt about a team going into the season since the 2011 juggernaut.

JayZee
12-10-2018, 05:48 PM
I subscribe to KenPom and I'd highly recommend it.

I don't think I'm giving up too much info here, and I'm also not sure how well the ratings hold up across the years, but this Duke team is still a good 2-3.5 efficiency margin points behind 4 other Duke teams of the 2000s. Of course 2 of those teams won NCs, one came an Emeka away from the NC game and the "best" one was one blown boozer call from the final four. Plus we have some time to improve...

House P
12-10-2018, 06:04 PM
Just to remind folks about the absolute monster that was the 1999 Blue Devils.

@ Ga Tech
@ St John

Those are the only two games from Jan 1 until the Final Four that 1999 Duke played that were not double digit victories. Duke actually went to OT against the Johnnies, winning by 4, and we only beat Ga Tech by 8.

Aside from those two, it was pretty much a massacre every time out.

SRS isn't a perfect metric. However, as far as I know, it is the only advanced team metric for which there is an online archive of pre-2002 ratings. Sportsreference.com lists the SRS rating of every college basketball team since 1950. The team with the highest SRS in college basketball history is the 1999 Duke team which was 34.80 points better than the average D1 team based on SRS.

How strong has Duke's performance been so far this year? According to SRS, the current Duke team is 33.69 points better than the average D1 team.

If Duke somehow keeps up at this pace*, it would end the year with the 3rd highest SRS rating of all time.

So its not surprising that advanced metrics currently find a fairly big separation between Duke and the second best team based on results through Dec 9.

In fact, I suspect that Duke would be even further ahead of the pack in KenPom if it weren't for the residual influence of pre-season ratings. Consider Duke and KenPom preseason #1 Kansas. Both teams have played 6 games vs top 100 KenPom teams

- Kansas has an average scoring margin of +9.0 vs Top 100 teams (including 2 top 25 teams)
- Duke has an average scoring margin of +19.6 vs Top 100 teams (including 4 top 25 teams)

According to KemPom Duke is "just" 4.16 points per 100 possessions better than Kansas. What do you think the KenPom margin would be if preseason ranking weren't included? I suspect the gap would be above 10 points per 100 possessions.


*Disclaimers about small sample sizes and regression to the mean apply here. For the record, I certainly don't expect Duke to keep up the current pace. After all, Duke has essentially blown out every non-top 10 team it has played so far. I don't think it is realistic to expect this to continue.

cato
12-10-2018, 07:17 PM
Unscientific, but correct, observation: if you were wandering around campus/the arena, devastated/searching for meaning, after the loss in either ‘86 or ‘99, that was the most painful one.

If you were there for both, the one that happened when you were closer to 18 hurts more.

KenPom’s 2019 tweaks prove this is true.

Dr. Rosenrosen
12-10-2018, 07:44 PM
Get a load of this joker... leaving every team that’s not played a true road game off his AP top 25 ballot.

https://amp.lansingstatejournal.com/amp/2264942002

He provides his email address in the article for anyone who disagrees and cares to share their opinion. I promptly emailed him asking on what planet one could consider the game against Kentucky, for all practical purposes, not to be a road game. Anyone who was there knows the pregame and first 10 minutes might as well have been taking place in Rupp.

JasonEvans
12-10-2018, 09:00 PM
In fact, I suspect that Duke would be even further ahead of the pack in KenPom if it weren't for the residual influence of pre-season ratings. Consider Duke and KenPom preseason #1 Kansas. Both teams have played 6 games vs top 100 KenPom teams

- Kansas has an average scoring margin of +9.0 vs Top 100 teams (including 2 top 25 teams)
- Duke has an average scoring margin of +19.6 vs Top 100 teams (including 4 top 25 teams)

According to KemPom Duke is "just" 4.16 points per 100 possessions better than Kansas. What do you think the KenPom margin would be if preseason ranking weren't included? I suspect the gap would be above 10 points per 100 possessions.

While Duke's preseason ranking is probably hurting them a little bit relative to Kansas, KenPom reduces the impact of the preseason bias with every game. While it is supposedly not completely gone until early January, it is probably getting pretty small at this point. Now, if Duke was something like the preseason #200 team, then maybe it would still be having a significant effect, but Duke was top 5. I can't imagine the small difference in Duke and Kansas in the preseason would change a 4.16 difference in PPP to 10+ PPP.

-Jason "as an aside, thanks for finding that SRS data on the 1999 team... historically great" Evans

House P
12-10-2018, 09:46 PM
While Duke's preseason ranking is probably hurting them a little bit relative to Kansas, KenPom reduces the impact of the preseason bias with every game. While it is supposedly not completely gone until early January, it is probably getting pretty small at this point. Now, if Duke was something like the preseason #200 team, then maybe it would still be having a significant effect, but Duke was top 5. I can't imagine the small difference in Duke and Kansas in the preseason would change a 4.16 difference in PPP to 10+ PPP.


KenPom doesn't tell us exactly how much preseason ratings affect his rankings at any given point in the year, so it is hard to know exactly what his ratings would look like if they were only based on games played.

However, KenPom bases his ratings on adjusted efficiency margin and you can do is get a fairly accurate estimate of the adjusted efficiency margin for each game. Kedsy basically does this in his "advanced stats" posts following each game.

If you do this for all 10 of Duke's game so far, you will find that they have an average adjusted efficiency margin of 39.2 for all games and 40.5 for games against top 100 competition. Kansas has an average adjusted efficiency margin of 22.7 for all games and 24.2 for games against top 100 competition. So this indicates that Duke KenPom rating may, in fact, be greater than 10 points per 100 possessions better than Kansas (39.2 vs 22.7) if it were based only on results so far this year (and not adjusted downwards for blowouts).

The tables below offer a detailed comparison of Duke and Kansas so far this year. The Adj EM column is essentially the KenPom rating for that game.

The Adj EM Rank column indicates what a team's KenPom ranking would be in the kept up that game's Adj EM for an entire season. So, one way to look at Duke's performance so far this year is that they have played 8 games like the #1 team in the country, 1 game like the #8 team, and one game like the #41 team.

Duke





Opponent
Opp Rank
Result
AdjOE
AdjDE
Adj EM
Adj EM Rank


Kentucky
19
W, 118-84
154.1
91.1
63.0
1


Army
262
W, 94-72
115.1
100.3
14.8
41


E Michigan
176
W, 84-46
111.4
65.2
46.2
1


San Diego St.
99
W, 90-64
125.4
83.1
42.3
1


Auburn
9
W, 78-72
123.3
90.6
32.7
1


Gonzaga
6
L, 89-87
128.8
104.4
24.4
8


Indiana
25
W, 90-69
122.7
83.6
39.0
1


Stetson
337
W, 113-49
123.0
66.8
56.3
1


Hartford
213
W, 84-54
109.6
77.5
32.1
1


Yale
92
W, 91-58
110.4
69.0
41.4
1





Kansas





Opponent
Opp
Rnk
Result
AdjOE
AdjDE
Adj EM
Adj EM Rank


Michigan St.
8
W, 92-87
119.9
90.0
29.8
2


Vermont
112
W, 84-68
114.5
92.6
21.9
12


Louisiana
135
W, 89-76
114.5
100.2
14.3
42


Marquette
36
W, 77-68
122.3
92.6
29.7
2


Tennessee
10
W, 87-81
118.1
87.9
30.2
2


Stanford
95
W, 90-84
120.4
110.4
10.0
65


Wofford
84
W, 72-47
104.5
66.6
37.9
1


NM St.
77
W, 63-60
101.6
94.0
7.6
93





I wish I had saved a copy of KenPom’s preseason rankings. That would help better understand how much they influence his current ratings.

uh_no
12-10-2018, 10:09 PM
I subscribe to KenPom and I'd highly recommend it.

I don't think I'm giving up too much info here, and I'm also not sure how well the ratings hold up across the years, but this Duke team is still a good 2-3.5 efficiency margin points behind 4 other Duke teams of the 2000s. Of course 2 of those teams won NCs, one came an Emeka away from the NC game and the "best" one was one blown boozer call from the final four. Plus we have some time to improve...
you can't really compare adjusted efficiencies across years.

COYS
12-10-2018, 10:20 PM
you can't really compare adjusted efficiencies across years.

If I understand JayZee correctly, he’s comparing how dominant Duke has been relative to this season’s opponents to how dominant duke has been to opponents in past seasons. Adjusted efficiency margin won’t tell us whether or not one team is actually better than another across seasons, but it can show how dominant that team is relative to the rest of college basketball in a given season, which seems to me to be as good of a measure of “dominance” as we’ve got.

tbyers11
12-10-2018, 10:38 PM
That's a huge stretch. For a so-far similarly exceptionally rated Duke team, to give you a comparison, duke has only a 72% chance of beating TT next week on a neutral court. I'd venture that uconn team was significantly better than this TT team, even if that duke team was somewhat better than this duke team...so an extra 8% chance of winning (and note the curve is non-linear...to go from 50->58% win is far easier than going from 70->78%) is a bit ridiculous.

That same St. Johns that took Duke to OT...lost to uconn in the big east final by 19. You understate Syracuse, who was in and out of the rankings. Uconn beat them by 22 and 21 points in the two rematches...including at the carrier dome and in MSG.

There's no doubt Duke was likely the better team...but 80% is absurd.

I don't think Jason's 80% is that absurd. The Vegas line for the Duke-UConn game in the 1999 final was Duke by 9.5. Vegas is usually within 1-2 of KP and a KP prediction of 8-9 points equates with 75-80% win probability. I'm not saying that Duke was 9.5 points better than UConn. We know that wasn't true given the results.

However, efficiency just looks at the numbers (also doesn't consider injuries). UConn not blowing out teams consistently throughout the season versus Duke winning nearly every game by large margins (including MD by 18 twice and UNC by 12, 20, and 23) would likely have led to a large efficiency margin. Efficiency does consider that UConn beat St Johns by 19 in the Big East final but it also considers that they only beat them by 4 earlier in the season and only beat a Seton Hall team (that didn't make the tourney) by 1 in the quarters.

Troublemaker
12-11-2018, 07:29 AM
I wish I had saved a copy of KenPom’s preseason rankings. That would help better understand how much they influence his current ratings.

Usually I'd recommend the Internet Archive for this: https://web.archive.org/web/*/kenpom.com

But it seems to be currently malfunctioning as no matter what date you click, it forwards to the snapshot taken on November 18th, which is about 3 games or so into the season for each team. Ideally, you'd have the rankings on the morning of November 6th, of course, as that was the first day of the season.

For kenpom subscribers, the equivalent of this is here: https://kenpom.com/archive.php?d=2018-11-06

Just change the date in the URL

HereBeforeCoachK
12-11-2018, 07:39 AM
I don't think Jason's 80% is that absurd. The Vegas line for the Duke-UConn game in the 1999 final was Duke by 9.5. Vegas is usually within 1-2 of KP and a KP prediction of 8-9 points equates with 75-80% win probability. I'm not saying that Duke was 9.5 points better than UConn. We know that wasn't true given the results.
.

I don't think the 80% is absurd either...I think that Duke team would beat that UConn team four out of five...just happens that the "fifth game", so to speak, was the one that took place on that day. And I do think they were 8-9 better than UConn, but those kind of upsets do happen...about one in five games.

CDu
12-11-2018, 07:54 AM
If I understand JayZee correctly, he’s comparing how dominant Duke has been relative to this season’s opponents to how dominant duke has been to opponents in past seasons. Adjusted efficiency margin won’t tell us whether or not one team is actually better than another across seasons, but it can show how dominant that team is relative to the rest of college basketball in a given season, which seems to me to be as good of a measure of “dominance” as we’ve got.

No, efficiency margin is offensive efficiency vs defensive efficiency.

We are ahead of all the 2000s in margin (relative to #2) of efficiency margin, but not in our own efficiency margin.

But as uh_no said, you can’t compare efficiency margins across years. You could compare relative difference in margins, but on that measure this team is best so far.

Dukehk
12-11-2018, 10:59 AM
Just to remind folks about the absolute monster that was the 1999 Blue Devils.

@ Ga Tech
@ St John

Those are the only two games from Jan 1 until the Final Four that 1999 Duke played that were not double digit victories. Duke actually went to OT against the Johnnies, winning by 4, and we only beat Ga Tech by 8.

Aside from those two, it was pretty much a massacre every time out. Maryland spent almost the entire season in the top 5 that year... we beat them by about 20 points every time we played them. Carolina was a top 10-15 teams all year... we also ripped them by 20 every time we played. We went undefeated in the ACC that year and set a record for victory margin. It was not uncommon for us to beat ACC teams by 30 or 40 points. We were on a 32 game winning streak when we fell to UConn.

As for UConn, they were really good in 1999, but nowhere near the level of dominance we saw from Duke. In Feburary they lost by 17 points (at home) to a decent but not great Syracuse team. They lost to Miami too. They had a nice win over a top 5 Stanford club by 11 points, but had a number of nailbiters against just so-so teams like Seton Hall and Providence.

I don't know what goes into KenPom's probability stats, but I suspect he would have had Duke more like 80% to win that national title game.

I remember so vividly watching that team when I was a kid. Pretty much the team that made me fall in love with Duke.

It seems like yesterday watching them massacre everyone on their way to the NT. Only to get my heartbroken by rip hamilton. I will never forget that absolutely disappointing gut wrenching feeling.

To this day I maintain that we would beat them 99 times out of 100.

1999 is by far the best team I have ever seen at the college level. We had lottery picks staying multiple years on that team. Elton Brand could have probably gone number 1 as a Freshman and was the best player in the country as a Soph. Corey Maggette was our sixth man for crying out loud. Langdon was an absolute ice cold (Alaskan) assassin...up until the final play of the year. William Avery was as athletic a point guard as you get, another lottery pick (too bad his career didn't pan out in the NBA). Battier was only in his second year but would have easily gone in the first round. Carawell was yet another athletic wing who would go on to become the ACC player of the year by his senior year. The bench was also so deep with Nate James, Burgess and even Domzalski. Nate was a lock down defender and Burgess/Domzalski were top 10 high school prospects and Mcdonalds all-americans.

That is the team that sets the barometer for every Duke team. Yes I know they didn't win the title, but to me that was the biggest collection of talent infused together as one unit for a single college season. 2018-2019 is shaping up to be great, but it still doesnt hold a candle to 1999 at the moment.

jv001
12-11-2018, 11:07 AM
I remember so vividly watching that team when I was a kid. Pretty much the team that made me fall in love with Duke.

It seems like yesterday watching them massacre everyone on their way to the NT. Only to get my heartbroken by rip hamilton. I will never forget that absolutely disappointing gut wrenching feeling.

To this day I maintain that we would beat them 99 times out of 100.

1999 is by far the best team I have ever seen at the college level. We had lottery picks staying multiple years on that team. Elton Brand could have probably gone number 1 as a Freshman and was the best player in the country as a Soph. Corey Maggette was our sixth man for crying out loud. Langdon was an absolute ice cold (Alaskan) assassin...up until the final play of the year. William Avery was as athletic a point guard as you get, another lottery pick (too bad his career didn't pan out in the NBA). Battier was only in his second year but would have easily gone in the first round. Carawell was yet another athletic wing who would go on to become the ACC player of the year by his senior year. The bench was also so deep with Nate James, Burgess and even Domzalski. Nate was a lock down defender and Burgess/Domzalski were top 10 high school prospects and Mcdonalds all-americans.

That is the team that sets the barometer for every Duke team. Yes I know they didn't win the title, but to me that was the biggest collection of talent infused together as one unit for a single college season. 2018-2019 is shaping up to be great, but it still doesnt hold a candle to 1999 at the moment.

The 1999 team was a great team and one of my all time favorites, but the 1992 team was my favorite and probably just as good as the 1999 team or better. That team had the added pressure of trying to repeat as NCAAT Champions. Since they accomplished that feat, I'll rank them higher than the 1999 team that failed to become Champions. Not saying it's a failure to finish 2nd, but being Champs is a lot better. GoDuke!

jimsumner
12-11-2018, 11:56 AM
The 1999 team was a great team and one of my all time favorites, but the 1992 team was my favorite and probably just as good as the 1999 team or better. That team had the added pressure of trying to repeat as NCAAT Champions. Since they accomplished that feat, I'll rank them higher than the 1999 team that failed to become Champions. Not saying it's a failure to finish 2nd, but being Champs is a lot better. GoDuke!

'92 v. '99?

Best of seven?

Probably go seven.

'92 has an edge in experience, starting two seniors (Laettner, Davis) and two juniors (Hurley, T. Hill), while '99 started three sophs, junior Carrawell and senior Langdon.

But '99 has a better bench, with Maggette, Burgess and James against Parks and Lang.

Future NBA success? Pretty equal it seems,

'99 had better stats but the ACC was a bit better in '92, methinks.

But ultimately, only one of these teams cut down the NCAA nets.

HereBeforeCoachK
12-11-2018, 12:04 PM
'92 v. '99?

Best of seven?

Probably go seven.

'92 has an edge in experience, starting two seniors (Laettner, Davis) and two juniors (Hurley, T. Hill), while '99 started three sophs, junior Carrawell and senior Langdon.

But '99 has a better bench, with Maggette, Burgess and James against Parks and Lang.

Future NBA success? Pretty equal it seems,

'99 had better stats but the ACC was a bit better in '92, methinks.

I would say to Dukehk....above...that the barometer was already set by 99...you might say 86...or you might say 92...but either way, it was pre set for 99.

But ultimately, only one of these teams cut down the NCAA nets.

Yep Yep...and not only did that team cut down the nets, they really did it twice with essentially the same team. That speaks to the incredible intangibles of Laettner, Hurley and the 91-92 teams...and it's the intangibles where I think 92 gets the big advantage over 99. To our buddy Dukehk above.....the barometer was set well before 99....maybe 86....certainly 91-92.....but either way, before 99.

COYS
12-11-2018, 12:37 PM
No, efficiency margin is offensive efficiency vs defensive efficiency.

We are ahead of all the 2000s in margin (relative to #2) of efficiency margin, but not in our own efficiency margin.

But as uh_no said, you can’t compare efficiency margins across years. You could compare relative difference in margins, but on that measure this team is best so far.

I guess I'm just missing something about what KenPom's adjusted efficiency margin means.

This year's team has a current adjusted efficiency margin of +30.58 meaning that (per KenPom's formula) Duke would outscore a perfectly average NCAA basketball team from the 2018-2019 season by 30.58 in 100 possessions (adj. offensive efficiency - adj. defensive efficiency = adj efficiency margin), right?

Why is it not valid to compare that margin (30.58) to previous seasons when attempting to determine how dominant one team was over its competition for a season compared to another? Is it not valid to say that the 18-19 Duke team has not yet reached the level of dominance of the '02 team because the '02 team would have been expected to beat the average competition from '01-'02 by 34.19 points in 100 possessions?

I get that KenPom's data resets every season so it is fruitless to compare the raw adj. efficiency numbers. In other words, we can't say that this year's team has a worse offense than last season's team because we have no way to know whether or not the competition is easier or harder this year compared to previous seasons. However, can't we say that this team is 30.58 points better per 100 possessions than the average competition from this season, which makes this team less dominant relative to the competition than than the '02 team (so far)? Am I missing something?

Ian
12-11-2018, 12:46 PM
Because we don't know how talent is distributed from year to year. Dominance is measured by how much better a team is than other top teams, not just average teams

If you're +30 in a year where there are many teams in the +25 to +30 range, then you're just barely ahead of the pack. If you're +30 in a year in where the are no teams who's +25, then you're a dominant team well ahead of the top competitors.

COYS
12-11-2018, 01:10 PM
Because we don't know how talent is distributed from year to year. Dominance is measured by how much better a team is than other top teams, not just average teams

If you're +30 in a year where there are many teams in the +25 to +30 range, then you're just barely ahead of the pack. If you're +30 in a year in where the are no teams who's +25, then you're a dominant team well ahead of the top competitors.

Right, I’ve acknowledged that we have no way of knowing how talent is distributed.

I think I see where my disconnect is. If we’re measuring dominance as being better than every single team by a wide margin, then measuring the efficiency gap between 1 and 2 makes sense.

MChambers
12-11-2018, 03:04 PM
Right, I’ve acknowledged that we have no way of knowing how talent is distributed.

I think I see where my disconnect is. If we’re measuring dominance as being better than every single team by a wide margin, then measuring the efficiency gap between 1 and 2 makes sense.

It's because of climate change?

Seriously, I agree with you.

NSDukeFan
12-11-2018, 03:46 PM
It's because of climate change?

Seriously, I agree with you.

It’s cold today. I’m pretty sure there’s no climate change.

Philadukie
12-12-2018, 08:00 AM
Here’s an interesting KenPom stat (or at least I find it interesting, but it’s probably meaningless at this point):

Our combined KenPom AdjO and AdjD ranking is currently 7 (2 for AdjO and 5 for AdjD).

Only two other teams since 2010 have finished a season with a combined ranking of 7 or better:

2010 Duke (6)
2015 Kentucky (7)

Next two closest would be:

2013 Louisville (8)
2016 Villanova (8)

The only other team with a combined ranking below 10 since 2010:

2012 Kentucky (9)

Right now our combined ranking puts us in elite company. It will be interesting to see if we can sustain this as the season unfolds. A combined ranking under 10 going into the tournament may be a sign of good things to come.

Clipsfan
12-12-2018, 09:51 AM
Yep Yep...and not only did that team cut down the nets, they really did it twice with essentially the same team. That speaks to the incredible intangibles of Laettner, Hurley and the 91-92 teams...and it's the intangibles where I think 92 gets the big advantage over 99. To our buddy Dukehk above....the barometer was set well before 99...maybe 86...certainly 91-92....but either way, before 99.

Revisionist history perhaps, we the '99 team was phenomenal and IIRC one shot in Alaska away from entering the NC undefeated. Off the top of my head, I think their average margin of victory was 26 points. I have no inside knowledge, but the rumors heavily abound that they were so cocky before the finals that they (at least some key players) blew off curfew the night before and weren't at their best that day. I agree that not winning the title was devastating as a student (a miserable night) but that shouldn't be the defining factor in how that team would do in a hypothetical matchup. It was an incredibly strong collection of talent and the most fun I've ever had watching a team in person.

uh_no
12-12-2018, 10:03 AM
Here’s an interesting KenPom stat (or at least I find it interesting, but it’s probably meaningless at this point):

Our combined KenPom AdjO and AdjD ranking is currently 7 (2 for AdjO and 5 for AdjD).

Only two other teams since 2010 have finished a season with a combined ranking of 7 or better:

2010 Duke (6)
2015 Kentucky (7)

Next two closest would be:

2013 Louisville (8)
2016 Villanova (8)

The only other team with a combined ranking below 10 since 2010:

2012 Kentucky (9)

Right now our combined ranking puts us in elite company. It will be interesting to see if we can sustain this as the season unfolds. A combined ranking under 10 going into the tournament may be a sign of good things to come.

I'm skeptical putting much weight behind it. There are, what, 7 other champions in the past decade who didn't meet this criteria? Teams that have very high combined efficiency ratings will be very good. Good teams are likely to win national championships. Taking the efficiency rating, making it far less powerful by taking the ranking is simply not going to be a more powerful indicator than the efficiencies themselves.

Duke is #1 in efficiency. They have a good chance to win the naional title. The fact that there happen to not be a few other teams clustered high in just offense or just defense to push our ranking down doesn't tell us anything new.

A better metric might be taking the square root of the difference between a teams offensive efficiency and the mean, and adding it to the same for defense. This would give you some sort of "metric" that rewards teams for being balanced on both sides of the ball. I'd STILL be skeptical of the usefulness of such a metric, since as is pointed out often here, there is no evidence that balance is an indicator of postseason success (or identically, success in general). If there were, then the KP premise that offense and defense can be analyzed independently becomes bogus, and the whole ranking system becomes invalid.

Troublemaker
12-12-2018, 12:06 PM
I listened to the first 10 minutes of KenPom's latest podcast (Episode 9): http://kenpom.libsyn.com/

Yes, he has a podcast cohosted with someone from The Athletic, and in those 10 minutes, KenPom weighed in on the subject of Duke's chances to become a "super-team"

It was interesting for sure.

KenPom thinks that since 1990, there have been 5 super-teams in men's college bball: 1991 UNLV, 1996 Kentucky, 1999 Duke, 2001 Duke, and 2015 Kentucky.

KenPom noted that only 2 of those 5 teams won a title, and he states that likewise, the odds will be against this year's Duke team winning it all -- that's just the nature of a single elimination tournament, he says.

But -- I've buried the lead -- KenPom DOES believe that Duke 2019 will be end up being a super-team alongside those other 5 teams.

Can't wait to finish the podcast because apparently the next part of the podcast is Kenpom interviewing / perhaps grilling the NCAA reps in charge of NET.

Duke79UNLV77
12-12-2018, 12:13 PM
I listened to the first 10 minutes of KenPom's latest podcast (Episode 9): http://kenpom.libsyn.com/

Yes, he has a podcast cohosted with someone from The Athletic, and in those 10 minutes, KenPom weighed in on the subject of Duke's chances to become a "super-team"

It was interesting for sure.

KenPom thinks that since 1990, there have been 5 super-teams in men's college bball: 1991 UNLV, 1996 Kentucky, 1999 Duke, 2001 Duke, and 2015 Kentucky.

KenPom noted that only 2 of those 5 teams won a title, and he states that likewise, the odds will be against this year's Duke team winning it all -- that's just the nature of a single elimination tournament, he says.

But -- I've buried the lead -- KenPom DOES believe that Duke 2019 will be end up being a super-team alongside those other 5 teams.

Can't wait to finish the podcast because apparently the next part of the podcast is Kenpom interviewing / perhaps grilling the NCAA reps in charge of NET.

Hopefully, this year we can push it up to half of the super teams winning national championships.

AGDukesky
12-12-2018, 01:06 PM
Well it seems making the Final Four is pretty likely if this becomes a Super Team. Interestingly, the two teams that eventually won the championship lost the most games prior to the NCAAs and played the most ranked teams during the entire season. Duke 2001 played a whopping 17 games against ranked teams by the time the trophy was lifted. UK 1996 and Duke 1999 both played 11. While the undefeated losers in the semifinals played the least at 9 (2015 UK) and 8 (1991 JNLV).

HereBeforeCoachK
12-12-2018, 02:01 PM
Revisionist history perhaps, we the '99 team was phenomenal and IIRC one shot in Alaska away from entering the NC undefeated. Off the top of my head, I think their average margin of victory was 26 points. I have no inside knowledge, but the rumors heavily abound that they were so cocky before the finals that they (at least some key players) blew off curfew the night before and weren't at their best that day. I agree that not winning the title was devastating as a student (a miserable night) but that shouldn't be the defining factor in how that team would do in a hypothetical matchup. It was an incredibly strong collection of talent and the most fun I've ever had watching a team in person.

I didn't interpret this discussion as being how they would do in a hypothetical match up. The 99 team would be quite a challenge for any team in that regard. But they didn't go through the grinder that 91-92 did, as first defending champions in a decade while arguably being the most popular and most hated team (simultaneously). I have no problem saying that the 99 team was far and away the best in the nation that year. But so was Duke 92...so much so that they peaked in December....and still had enough to win it all.

And without his injury, Grant Hill would've dwarfed everybody on the 99 teams NBA career combined IMO.

House P
12-12-2018, 02:46 PM
KenPom doesn't tell us exactly how much preseason ratings affect his rankings at any given point in the year, so it is hard to know exactly what his ratings would look like if they were only based on games played.

I wish I had saved a copy of KenPom’s preseason rankings. That would help better understand how much they influence his current ratings.


Thanks to Troublemaker's guidance on finding preseason KenPom ratings, I made a very rough estimate of how much his current ratings are still influenced by his preseason ratings. A short description of my methodology is below, but here is the takeaway.

It appears that approximately 50% of KenPom's current ratings are still based on his preseason ratings.

Based on this, you can estimate what a team's KenPom rating would be without the influence of pre-season ratings using the following formula:

Rating without preseason component = (Current KenPom Rating - Preseason KenPom Rating) * 2 + Preseason KenPom Rating

For Duke this would be,

Current KenPom rating = 30.57
Preseason KenPom rating = 25.51
Rating without preseason component = 35.62

So, as far as Duke is concerned, KenPom's current ratings (which are already outstanding) likely underestimate how dominate Duke has been so far this year. As such, it is not surprising the KenPom sees Duke as a potential "Superteam".

Here is a very rough estimate of what KenPom's top 50 might look like if the preseason ratings were removed. Duke and Michigan are the top 2 teams using this metric (with quite a bit of separation between the rest of the field). Keep in mind, that there is good reason to include a preseason component at this point in the season to account for likely regression to the mean for outliers. Otherwise, you have some highly controversial ratings which probably won't hold up as more games are played - such as Virginia Tech at #3, Kansas at #14, Utah St at #15, San Francisco at #18, Kentucky at #46, and Villanova at #57.






AdjEM
Rank
AdjOE
Rank
Adj DE
Rank
Current AdjEM – Preseason AdjEM


Duke
35.64
1
123.04
1
87.47
3
5.07


Michigan
34.37
2
118.21
7
83.90
1
8.21


Virginia Tech
28.35
3
119.94
4
91.66
18
6.08


Michigan St.
28.19
4
119.85
5
91.65
17
3.9


Auburn
27.76
5
118.04
8
90.27
9
4.07


Gonzaga
27.43
6
122.95
2
95.55
48
2.1


Virginia
26.99
7
116.15
12
89.15
6
0.7


Nevada
26.83
8
120.70
3
93.81
30
1.19


Texas Tech
25.92
9
112.28
32
86.43
2
4.15


Tennessee
25.11
10
116.09
13
90.92
14
1.97


Ohio St.
24.23
11
114.25
23
90.06
8
5.29


Wisconsin
23.81
12
114.68
18
90.82
12
2.36


North Carolina
23.71
13
118.58
6
94.83
41
-0.85


Kansas
23.68
14
115.41
16
91.80
19
-2.79


Utah St.
23.02
15
114.38
22
91.42
16
11.56


Buffalo
22.95
16
116.16
11
93.14
27
6.52


Nebraska
22.95
16
116.01
14
93.10
26
4.21


San Francisco
22.56
18
111.61
37
89.00
5
10.01


Oklahoma
20.80
19
111.69
36
90.92
13
4.37


Purdue
20.65
20
116.27
9
95.63
51
0.54


Iowa St.
20.47
21
114.63
19
94.17
36
1.26


Cincinnati
20.00
22
110.80
51
90.81
11
3.19


Florida St.
19.96
23
112.14
33
92.16
20
-0.12


North Carolina St.
19.52
24
116.24
10
96.77
63
2.27


Louisville
18.72
25
113.61
26
94.80
40
3.81


Butler
18.48
26
113.71
24
95.20
45
1.01


Houston
18.39
27
111.45
41
93.06
25
2.49


Mississippi St.
18.38
28
112.36
31
94.05
34
0.11


Indiana
18.23
29
111.33
43
93.18
28
0.65


Maryland
17.78
30
112.07
34
94.34
38
0.53


Florida
17.65
31
108.65
75
90.95
15
-0.7


TCU
17.43
32
111.28
45
93.83
31
0.75


UCF
17.02
33
109.71
64
92.70
22
3.41


Arizona
16.96
34
110.88
50
93.93
33
3.75


Arizona St.
16.29
35
111.12
48
94.79
39
2.23


Oregon
15.99
36
111.91
35
95.90
53
-0.59


Kansas St.
15.45
37
105.33
131
89.87
7
-2.54


Marquette
15.19
38
111.32
44
96.08
55
-0.76


Lipscomb
15.13
39
111.35
42
96.26
57
6.09


Northwestern
15.12
40
108.67
74
93.53
29
1.74


Syracuse
15.06
41
108.90
71
93.90
32
-4.2


Saint Mary's
14.88
42
115.37
17
100.54
119
2.73


Arkansas
14.88
42
110.19
58
95.32
47
1.97


Iowa
14.84
44
114.58
20
99.73
101
-0.25


Fresno St.
14.68
45
109.81
62
95.10
43
4.81


Kentucky
14.45
46
112.87
28
98.43
83
-5.06


Mississippi
14.12
47
113.03
27
98.97
88
3.66


Creighton
14.07
48
115.73
15
101.67
140
-0.94


Colorado
13.96
49
109.51
65
95.60
50
3.41


UCLA
13.92
50
110.55
54
96.65
60
0.5





METHODOLOGY

Here is a overview of the steps I took to estimate the preseason component of the current KenPom ratings. If anyone would like more details, I could provide a more complete description when I get some free time.

1) Identify the 50 teams with the greatest difference between their current KenPom rating and their preseason rating.

2) Calculate the game-by-game Adjusted EM for each of these 50 teams. The average of the game-by-game Adjusted EM is the team's Estimated Adjusted EM with no contribution from the preseason rating.

3) Determine how much weight the preseason component would have to have in order to end up with the current rating based on the following formula: Current AdjEM = Preseason AdjEM * X + Estimated AdjEM * (1-X)

The average value of X was 0.51 with a standard deviation of 0.08.

The value of X seems to decrease based on the number of games played (0.53 for 8 GP and 0.45 for 10 GP) and the number of blowouts (more blowouts tended to increase X). It is possible that a "more accurate" formula would takes games played and blowouts into account, but it is probably close enough to just estimate the value to be 0.5 for everyone at this point in the year.

uh_no
12-12-2018, 05:17 PM
Thanks to Troublemaker's guidance on finding preseason KenPom ratings, I made a very rough estimate of how much his current ratings are still influenced by his preseason ratings. A short description of my methodology is below, but here is the takeaway.

It appears that approximately 50% of KenPom's current ratings are still based on his preseason ratings.

Based on this, you can estimate what a team's KenPom rating would be without the influence of pre-season ratings using the following formula:

Rating without preseason component = (Current KenPom Rating - Preseason KenPom Rating) * 2 + Preseason KenPom Rating

For Duke this would be,

Current KenPom rating = 30.57
Preseason KenPom rating = 25.51
Rating without preseason component = 35.62

So, as far as Duke is concerned, KenPom's current ratings (which are already outstanding) likely underestimate how dominate Duke has been so far this year. As such, it is not surprising the KenPom sees Duke as a potential "Superteam".

Here is a very rough estimate of what KenPom's top 50 might look like if the preseason ratings were removed. Duke and Michigan are the top 2 teams using this metric (with quite a bit of separation between the rest of the field). Keep in mind, that there is good reason to include a preseason component at this point in the season to account for likely regression to the mean for outliers. Otherwise, you have some highly controversial ratings which probably won't hold up as more games are played - such as Virginia Tech at #3, Kansas at #14, Utah St at #15, San Francisco at #18, Kentucky at #46, and Villanova at #57.






AdjEM
Rank
AdjOE
Rank
Adj DE
Rank
Current AdjEM – Preseason AdjEM


Duke
35.64
1
123.04
1
87.47
3
5.07


Michigan
34.37
2
118.21
7
83.90
1
8.21


Virginia Tech
28.35
3
119.94
4
91.66
18
6.08


Michigan St.
28.19
4
119.85
5
91.65
17
3.9


Auburn
27.76
5
118.04
8
90.27
9
4.07


Gonzaga
27.43
6
122.95
2
95.55
48
2.1


Virginia
26.99
7
116.15
12
89.15
6
0.7


Nevada
26.83
8
120.70
3
93.81
30
1.19


Texas Tech
25.92
9
112.28
32
86.43
2
4.15


Tennessee
25.11
10
116.09
13
90.92
14
1.97


Ohio St.
24.23
11
114.25
23
90.06
8
5.29


Wisconsin
23.81
12
114.68
18
90.82
12
2.36


North Carolina
23.71
13
118.58
6
94.83
41
-0.85


Kansas
23.68
14
115.41
16
91.80
19
-2.79


Utah St.
23.02
15
114.38
22
91.42
16
11.56


Buffalo
22.95
16
116.16
11
93.14
27
6.52


Nebraska
22.95
16
116.01
14
93.10
26
4.21


San Francisco
22.56
18
111.61
37
89.00
5
10.01


Oklahoma
20.80
19
111.69
36
90.92
13
4.37


Purdue
20.65
20
116.27
9
95.63
51
0.54


Iowa St.
20.47
21
114.63
19
94.17
36
1.26


Cincinnati
20.00
22
110.80
51
90.81
11
3.19


Florida St.
19.96
23
112.14
33
92.16
20
-0.12


North Carolina St.
19.52
24
116.24
10
96.77
63
2.27


Louisville
18.72
25
113.61
26
94.80
40
3.81


Butler
18.48
26
113.71
24
95.20
45
1.01


Houston
18.39
27
111.45
41
93.06
25
2.49


Mississippi St.
18.38
28
112.36
31
94.05
34
0.11


Indiana
18.23
29
111.33
43
93.18
28
0.65


Maryland
17.78
30
112.07
34
94.34
38
0.53


Florida
17.65
31
108.65
75
90.95
15
-0.7


TCU
17.43
32
111.28
45
93.83
31
0.75


UCF
17.02
33
109.71
64
92.70
22
3.41


Arizona
16.96
34
110.88
50
93.93
33
3.75


Arizona St.
16.29
35
111.12
48
94.79
39
2.23


Oregon
15.99
36
111.91
35
95.90
53
-0.59


Kansas St.
15.45
37
105.33
131
89.87
7
-2.54


Marquette
15.19
38
111.32
44
96.08
55
-0.76


Lipscomb
15.13
39
111.35
42
96.26
57
6.09


Northwestern
15.12
40
108.67
74
93.53
29
1.74


Syracuse
15.06
41
108.90
71
93.90
32
-4.2


Saint Mary's
14.88
42
115.37
17
100.54
119
2.73


Arkansas
14.88
42
110.19
58
95.32
47
1.97


Iowa
14.84
44
114.58
20
99.73
101
-0.25


Fresno St.
14.68
45
109.81
62
95.10
43
4.81


Kentucky
14.45
46
112.87
28
98.43
83
-5.06


Mississippi
14.12
47
113.03
27
98.97
88
3.66


Creighton
14.07
48
115.73
15
101.67
140
-0.94


Colorado
13.96
49
109.51
65
95.60
50
3.41


UCLA
13.92
50
110.55
54
96.65
60
0.5





METHODOLOGY

Here is a overview of the steps I took to estimate the preseason component of the current KenPom ratings. If anyone would like more details, I could provide a more complete description when I get some free time.

1) Identify the 50 teams with the greatest difference between their current KenPom rating and their preseason rating.

2) Calculate the game-by-game Adjusted EM for each of these 50 teams. The average of the game-by-game Adjusted EM is the team's Estimated Adjusted EM with no contribution from the preseason rating.

3) Determine how much weight the preseason component would have to have in order to end up with the current rating based on the following formula: Current AdjEM = Preseason AdjEM * X + Estimated AdjEM * (1-X)

The average value of X was 0.51 with a standard deviation of 0.08.

The value of X seems to decrease based on the number of games played (0.53 for 8 GP and 0.45 for 10 GP) and the number of blowouts (more blowouts tended to increase X). It is possible that a "more accurate" formula would takes games played and blowouts into account, but it is probably close enough to just estimate the value to be 0.5 for everyone at this point in the year.

I think there are a couple things that may cloud the analysis slightly.

1) The graph simply isn't well-connected enough yet. The same reason the NET rankings were bogus a couple weeks ago, and still had some head scratchers last week is evidence of that. That's a trend that you would expect to be non-linear and unpredictable. If you consider splitting the country into two halves, and those two halves only play eachother. We get a bunch of data to rank the teams, and do our best to come up with a relative ranking of the teams. Then one day we have "inter league play" and the teams only play teams from the other half. We find that one half was way overrated vs the other half and the system makes a huge adjustment to all those teams based on that new data. This is what happens every day early season in KP as groups of teams become better linked as new teams play eachother. Until those games actually happen, you can't predict which way each teams rankings will have to be adjusted.

In short, this means that trying to extrapolate a trend is not necessarily a sound method. You would expect this effect to be somewhat modelable by a random walk, which is gaussian. This means that even if a team is trending upwards after a few trials, trying to extrapolate that out will lead to an overestimation of the future trend. (hot hand fallacy...which is amusingly a real effect in sports)

2) the estimation of the preseason component is shaky. Kenpom's game weights are non-linear. Your previous game is worth more than your next previous is worth more...etc. Your method simply takes a flat average. Without knowing the distribution a bit more formally, one could be off a significant amount. If I had to wager a guess, I'd say IF this number is wrong, it is OVERSTATING the current weight of preseason rankings.

So all in all, I expect that while duke's current KP is an underestimation of their current performance, I also expect that your extrapolation is a significant overestimation, and without actually executing the model sans preseason data, it's difficult to know by how much.

Major effort sporks. made me think.

-jk
12-12-2018, 05:31 PM
Thanks to Troublemaker's guidance on finding preseason KenPom ratings, I made a very rough estimate of how much his current ratings are still influenced by his preseason ratings. A short description of my methodology is below, but here is the takeaway.

It appears that approximately 50% of KenPom's current ratings are still based on his preseason ratings.

Based on this, you can estimate what a team's KenPom rating would be without the influence of pre-season ratings using the following formula:

Rating without preseason component = (Current KenPom Rating - Preseason KenPom Rating) * 2 + Preseason KenPom Rating

For Duke this would be,

Current KenPom rating = 30.57
Preseason KenPom rating = 25.51
Rating without preseason component = 35.62

So, as far as Duke is concerned, KenPom's current ratings (which are already outstanding) likely underestimate how dominate Duke has been so far this year. As such, it is not surprising the KenPom sees Duke as a potential "Superteam".

Here is a very rough estimate of what KenPom's top 50 might look like if the preseason ratings were removed. Duke and Michigan are the top 2 teams using this metric (with quite a bit of separation between the rest of the field). Keep in mind, that there is good reason to include a preseason component at this point in the season to account for likely regression to the mean for outliers. Otherwise, you have some highly controversial ratings which probably won't hold up as more games are played - such as Virginia Tech at #3, Kansas at #14, Utah St at #15, San Francisco at #18, Kentucky at #46, and Villanova at #57.






AdjEM
Rank
AdjOE
Rank
Adj DE
Rank
Current AdjEM – Preseason AdjEM


Duke
35.64
1
123.04
1
87.47
3
5.07


Michigan
34.37
2
118.21
7
83.90
1
8.21


Virginia Tech
28.35
3
119.94
4
91.66
18
6.08


Michigan St.
28.19
4
119.85
5
91.65
17
3.9


Auburn
27.76
5
118.04
8
90.27
9
4.07


Gonzaga
27.43
6
122.95
2
95.55
48
2.1


Virginia
26.99
7
116.15
12
89.15
6
0.7


Nevada
26.83
8
120.70
3
93.81
30
1.19


Texas Tech
25.92
9
112.28
32
86.43
2
4.15


Tennessee
25.11
10
116.09
13
90.92
14
1.97


Ohio St.
24.23
11
114.25
23
90.06
8
5.29


Wisconsin
23.81
12
114.68
18
90.82
12
2.36


North Carolina
23.71
13
118.58
6
94.83
41
-0.85


Kansas
23.68
14
115.41
16
91.80
19
-2.79


Utah St.
23.02
15
114.38
22
91.42
16
11.56


Buffalo
22.95
16
116.16
11
93.14
27
6.52


Nebraska
22.95
16
116.01
14
93.10
26
4.21


San Francisco
22.56
18
111.61
37
89.00
5
10.01


Oklahoma
20.80
19
111.69
36
90.92
13
4.37


Purdue
20.65
20
116.27
9
95.63
51
0.54


Iowa St.
20.47
21
114.63
19
94.17
36
1.26


Cincinnati
20.00
22
110.80
51
90.81
11
3.19


Florida St.
19.96
23
112.14
33
92.16
20
-0.12


North Carolina St.
19.52
24
116.24
10
96.77
63
2.27


Louisville
18.72
25
113.61
26
94.80
40
3.81


Butler
18.48
26
113.71
24
95.20
45
1.01


Houston
18.39
27
111.45
41
93.06
25
2.49


Mississippi St.
18.38
28
112.36
31
94.05
34
0.11


Indiana
18.23
29
111.33
43
93.18
28
0.65


Maryland
17.78
30
112.07
34
94.34
38
0.53


Florida
17.65
31
108.65
75
90.95
15
-0.7


TCU
17.43
32
111.28
45
93.83
31
0.75


UCF
17.02
33
109.71
64
92.70
22
3.41


Arizona
16.96
34
110.88
50
93.93
33
3.75


Arizona St.
16.29
35
111.12
48
94.79
39
2.23


Oregon
15.99
36
111.91
35
95.90
53
-0.59


Kansas St.
15.45
37
105.33
131
89.87
7
-2.54


Marquette
15.19
38
111.32
44
96.08
55
-0.76


Lipscomb
15.13
39
111.35
42
96.26
57
6.09


Northwestern
15.12
40
108.67
74
93.53
29
1.74


Syracuse
15.06
41
108.90
71
93.90
32
-4.2


Saint Mary's
14.88
42
115.37
17
100.54
119
2.73


Arkansas
14.88
42
110.19
58
95.32
47
1.97


Iowa
14.84
44
114.58
20
99.73
101
-0.25


Fresno St.
14.68
45
109.81
62
95.10
43
4.81


Kentucky
14.45
46
112.87
28
98.43
83
-5.06


Mississippi
14.12
47
113.03
27
98.97
88
3.66


Creighton
14.07
48
115.73
15
101.67
140
-0.94


Colorado
13.96
49
109.51
65
95.60
50
3.41


UCLA
13.92
50
110.55
54
96.65
60
0.5





METHODOLOGY

Here is a overview of the steps I took to estimate the preseason component of the current KenPom ratings. If anyone would like more details, I could provide a more complete description when I get some free time.

1) Identify the 50 teams with the greatest difference between their current KenPom rating and their preseason rating.

2) Calculate the game-by-game Adjusted EM for each of these 50 teams. The average of the game-by-game Adjusted EM is the team's Estimated Adjusted EM with no contribution from the preseason rating.

3) Determine how much weight the preseason component would have to have in order to end up with the current rating based on the following formula: Current AdjEM = Preseason AdjEM * X + Estimated AdjEM * (1-X)

The average value of X was 0.51 with a standard deviation of 0.08.

The value of X seems to decrease based on the number of games played (0.53 for 8 GP and 0.45 for 10 GP) and the number of blowouts (more blowouts tended to increase X). It is possible that a "more accurate" formula would takes games played and blowouts into account, but it is probably close enough to just estimate the value to be 0.5 for everyone at this point in the year.

Nice crunching. Kenpom will be "well connected" soon enough. Well, except that conferences get way too incestuous after the new year. We need Sagegrouse's (wise or cranky; both work) plan for a week of inter-conference play sometime in Feb to help rebalance everyone.

-jk

JayZee
12-12-2018, 05:53 PM
Here’s an interesting KenPom stat (or at least I find it interesting, but it’s probably meaningless at this point):

Our combined KenPom AdjO and AdjD ranking is currently 7 (2 for AdjO and 5 for AdjD).

Only two other teams since 2010 have finished a season with a combined ranking of 7 or better:

2010 Duke (6)
2015 Kentucky (7)

Next two closest would be:

2013 Louisville (8)
2016 Villanova (8)

The only other team with a combined ranking below 10 since 2010:

2012 Kentucky (9)

Right now our combined ranking puts us in elite company. It will be interesting to see if we can sustain this as the season unfolds. A combined ranking under 10 going into the tournament may be a sign of good things to come.

Duke 2002 - #1 Offense, #1 Defense, so a 2. And 4 efficiency points better than the #2 team. Maybe one of the most under-rated great Coach K Duke teams alongside 2004.

House P
12-12-2018, 06:12 PM
I think there are a couple things that may cloud the analysis slightly.

2) the estimation of the preseason component is shaky. Kenpom's game weights are non-linear. Your previous game is worth more than your next previous is worth more...etc. Your method simply takes a flat average. Without knowing the distribution a bit more formally, one could be off a significant amount. If I had to wager a guess, I'd say IF this number is wrong, it is OVERSTATING the current weight of preseason rankings.

So all in all, I expect that while duke's current KP is an underestimation of their current performance, I also expect that your extrapolation is a significant overestimation, and without actually executing the model sans preseason data, it's difficult to know by how much.


Thanks for the thought provoking comments.

Do you know of a reference which describes how rapidly KenPom discounts previous games as the season progresses? The closest thing I can find is in the T-Rank FAQ on Barttorvik.com (http://adamcwisports.blogspot.com/p/every-possession-counts.html). In T-Rank, which is similar to (but not the same as) KenPom, all games within the previous 40 days count as 100%. He then discounts the games at 1% per day until 80 days, at which point they count 60%. If KenPom is using a similar approach, no games are currently being discounted.

As far as preseason rankings, The T-Rank FAQ mentions that "as with Kenpom, there is also a preseason component that is phased out once a team has played 13 adjusted games (since not all games count for 100% of a game, it typically sticks around for 15 or 16 games). Because, the T-Rank ratings seems to show more movement vs preseason rankings than KenPom*, I wouldn't be surprised if T-Rank eliminates pre-season influence more rapidly than KemPom. The closest thing I can find from KenPom is this link (https://kenpom.com/blog/preseason-ratings-why-weight/)where he talks about the value of including preseason influence beyond Dec 24 (and hints that Nate Silver thinks you should include it through the end of the season).

If KenPom uses the same approach to phasing out pre-season rankings as T-Rank, I would expect that Duke has probably played somewhere between 7 and 8 "adjusted games" due to all the blowouts. This would imply that the preseason rating might still be contributing 40% or more to Duke's current rating (if the phase out is linear, which it may not be).

I think I may have found a way to get T-Rank ratings without preseason influence. Based on the filters used in this link (http://www.barttorvik.com/?year=2019&sort=&lastx=0&hteam=&conlimit=All&state=All&begin=20181106&end=20181211&top=0&quad=4&venue=All&type=All&mingames=0#), Duke's Adj EM from Nov 6- Dec 11 is 37.2 (Adj OE = 122.4, Adj DE = 85.2).

devildeac
12-12-2018, 06:18 PM
Duke 2002 - #1 Offense, #1 Defense, so a 2. And 4 efficiency points better than the #2 team. Maybe one of the most under-rated great Coach K Duke teams alongside 2004.


Yep, quite an assortment of talent.

And yes, Boozer was still fouled :mad:.

HereBeforeCoachK
12-12-2018, 06:26 PM
Yep, quite an assortment of talent.

And yes, Boozer was still fouled :mad:.

...and yes, we will never forget Bruce Eggs Benedict and his pathetic officiating all night in that game...

DavidBenAkiva
12-12-2018, 08:37 PM
Duke 2002 - #1 Offense, #1 Defense, so a 2. And 4 efficiency points better than the #2 team. Maybe one of the most under-rated great Coach K Duke teams alongside 2004.

I had it so good. My freshmen year was in 2000-01, so really started paying attention to Duke basketball in '99 and then was on campus through the 2006 season (took a couple years off to grow up). I was treated to some of the best teams and players college basketball has ever seen.

When all is said and done, there will be several peaks and valleys to Coach K's career.

1986-94 - reaching the pinnacle, the closest any team has gotten to John Wooden's UCLA teams
1999-2006 - back with a vengeance, proving the first act was no accident
2010-13 - mini peak! the last of the 4-year college teams
2015-present - master class is in session, now listen up all you one-and-doners

This Duke team has a chance to enter the rare air of some of the best and it's so exciting to see it happen. I just hope we can stay healthy.

uh_no
12-13-2018, 12:39 AM
Thanks for the thought provoking comments.

Do you know of a reference which describes how rapidly KenPom discounts previous games as the season progresses?

I wish!

unfortunately, KP has become (seeming to ME anyway...could be a bias) somewhat more opaque to his methods over time. While he justifies major changes, it doesn't seem he talks AS much about the nitty gritty as he used to. Perhaps it's because he just doesn't blog as much....or perhaps it's because the system doesn't change as much....or perhaps it's because he's successfully monetized it to the degree that he has a significant interest in not divulging the nitty gritty....

Either way, as far as I know, there is SOME factor weighting games by recency, and there is SOME time after which preseason rankings are discarded, usually in late january, if I recall.

Philadukie
12-13-2018, 08:40 AM
I'm skeptical putting much weight behind it. There are, what, 7 other champions in the past decade who didn't meet this criteria? Teams that have very high combined efficiency ratings will be very good. Good teams are likely to win national championships. Taking the efficiency rating, making it far less powerful by taking the ranking is simply not going to be a more powerful indicator than the efficiencies themselves.

Duke is #1 in efficiency. They have a good chance to win the naional title. The fact that there happen to not be a few other teams clustered high in just offense or just defense to push our ranking down doesn't tell us anything new.

A better metric might be taking the square root of the difference between a teams offensive efficiency and the mean, and adding it to the same for defense. This would give you some sort of "metric" that rewards teams for being balanced on both sides of the ball. I'd STILL be skeptical of the usefulness of such a metric, since as is pointed out often here, there is no evidence that balance is an indicator of postseason success (or identically, success in general). If there were, then the KP premise that offense and defense can be analyzed independently becomes bogus, and the whole ranking system becomes invalid.

Here is a list of all of the teams since 2002 with a combined AdjO and AdjD ranking of under 10:

2016 Nova (8) - NC
2015 Kentucky (7) - FF
2013 Louisville (8) - NC
2012 Kentucky (9) - NC
2010 Duke (6) - NC
2008 Kansas (3) - NC
2008 Memphis (8) - FF/CG
2007 UNC (7) - E8
2005 UNC (5) - NC
2005 Illinois (7) - FF/CG
2004 Duke (5) - FF
2003 Kentucky (9) - E8
2002 Duke (1) - S16

Of the 13 teams that have had combined AdjO and AdjD rankings of less than 10 since 2002, 6 have been National Champions (46%), 8 have played in the championship game (62%), and 10 have been in the Final Four (77%). My point wasn't that other teams with higher combined scores can't be National Champions, my point was that if you see a team with a combined ranking of less than 10, it's a pretty good rule of thumb that that team will probably make the Final Four and may even be your National Champion.

I'm just pointing out what I think is a useful unintended heuristic available in KenPom's data for identifying more likely than not FF teams. To your point, it's not exclusive criteria. Other teams with higher combined rankings still make the FF and win the NC (more on that in a second), but teams with combined rankings under 10 are in the Final Four 77% of the time.

Right now, Duke has the combined ranking of a team that has an excellent chance of making at least a FF (and even NC game). It's early, and we'll see how things unfold, but if they're under 10 at the time of the tournament, you can feel pretty good about their chances (no other team is under 10 currently).

I also looked at all of the National Champion teams since 2002. I won't list them here, but all but three of the 17 teams (UConn - '14, '11, Cuse - '03) had a combined ranking of 20 or below (82%). Again, a combined ranking under 20 isn't sufficiently dispositive (many teams with scores under 20, of course, aren't National Champions), but it at least gives you a pretty good indicator of what a National Champion should look like.

The one major flaw in all of this is that I'm looking at post-tournament data. I'm sure not all of the NC winners were under 20 at the start of the tournament. But, when it comes to identifying potential FF and even NC teams, I still think it's a helpful rule of thumb: if you see a team that's ranked, say, number 2 on KenPom, but has a combined ranking of over 20 (like Virginia last year -31), that's probably not your National Champion. And if you see a team with a combined ranking of under 10, it's a pretty good chance that they will at least make the FF.

In that sense, I actually do think it adds more utility than just the efficiency rankings themselves. There are plenty of teams each year in the top 10 efficiency rankings that have relatively higher combined AdjO and AdjD rankings than teams they're ranked above.

uh_no
12-13-2018, 09:23 AM
My point wasn't that other teams with higher combined scores can't be National Champions, my point was that if you see a team with a combined ranking of less than 10, it's a pretty good rule of thumb that that team will probably make the Final Four and may even be your National Champion.

Right....but all you're saying is "teams that are really good are likely to do well in the tournament"

I'm not arguing against the truthfulness of the correlation, I'm just skeptical of its utility. If you had a march madness contest where you picked how two teams would match up based on actual KP ratings, and another where you picked teams based on who had a lower combined sum of O + D rating, I would bet at least 1 pie, maybe 2 that the actual ratings perform better. If you want, remind me in march, and I'll track it all.

Relevant XKCD (mildly nsfw text):

https://xkcd.com/1138/

Philadukie
12-13-2018, 10:46 AM
Right...but all you're saying is "teams that are really good are likely to do well in the tournament"

I'm not arguing against the truthfulness of the correlation, I'm just skeptical of its utility. If you had a march madness contest where you picked how two teams would match up based on actual KP ratings, and another where you picked teams based on who had a lower combined sum of O + D rating, I would bet at least 1 pie, maybe 2 that the actual ratings perform better. If you want, remind me in march, and I'll track it all.

Relevant XKCD (mildly nsfw text):

https://xkcd.com/1138/

That’s not what I’m saying though. I’m saying two very specific things (and not generally “good teams do well in the tournament”) that offer expanded utility beyond the overall rankings:

1) if a team has a combined AdjO and AdjD of less than 10, based on the last 17 years, there’s a 46% chance they won a N.C. and 77% chance they were in the FF.

2) of the last 17 NC teams, 83% had a combined AdjO and AdjD ranking of 20 or fewer.

With post hoc ergo propter reasoning aside (because I’m looking at post-tournament data), this is telling me that if a team has a combined ranking of less than 10 going into the tournament, there are very good chances (relative to the strength of the rest of the field) that this team will be a FF team and even a NC.

I’ll put a couple pie bets on this :).

House P
12-13-2018, 11:41 AM
Here is a list of all of the teams since 2002 with a combined AdjO and AdjD ranking of under 10:

2016 Nova (8) - NC
2015 Kentucky (7) - FF
2013 Louisville (8) - NC
2012 Kentucky (9) - NC
2010 Duke (6) - NC
2008 Kansas (3) - NC
2008 Memphis (8) - FF/CG
2007 UNC (7) - E8
2005 UNC (5) - NC
2005 Illinois (7) - FF/CG
2004 Duke (5) - FF
2003 Kentucky (9) - E8
2002 Duke (1) - S16

Of the 13 teams that have had combined AdjO and AdjD rankings of less than 10 since 2002, 6 have been National Champions (46%), 8 have played in the championship game (62%), and 10 have been in the Final Four (77%). My point wasn't that other teams with higher combined scores can't be National Champions, my point was that if you see a team with a combined ranking of less than 10, it's a pretty good rule of thumb that that team will probably make the Final Four and may even be your National Champion.

The one major flaw in all of this is that I'm looking at post-tournament data. I'm sure not all of the NC winners were under 20 at the start of the tournament. But, when it comes to identifying potential FF and even NC teams, I still think it's a helpful rule of thumb: if you see a team that's ranked, say, number 2 on KenPom, but has a combined ranking of over 20 (like Virginia last year -31), that's probably not your National Champion. And if you see a team with a combined ranking of under 10, it's a pretty good chance that they will at least make the FF.


As you note, the use of post-tourney rankings complicates the analysis. After all, every National Championship team ended the season on a 6 game winning streak against top quality competition, so their final rankings are almost certainly higher than their pre-tourney ranking.

In fact, two of the National Champions on your list (2016 Nova and 2013 Louisville) would not meet your criteria if you used pre-tourney rankings. In addition, two teams which flamed out in the round of 32 (2010 Kansas and 2002 Cincy) would have met your criteria if pre-tourney rankings were used.

For what it is worth, the table below shows the pre- and post- tourney ranking (Adj OE Rank + Adj DE Rank) of each of the teams you listed and two which would have met your criteria based on pre-tourney rankings. It turns out that teams which meet your criteria based on pre-tourney rankings have a couple other things in common beyond meeting your criteria: a) they received a 1 seed, and b) they entered the tourney either #1 overall (8 teams) or #2 overall (3 teams) in KenPom's pre-tourney rankings.

To convince me that there is something especially predictive of your criteria, I would like to see if teams which met your criteria performed significantly better than a similar cohort of teams which didn't meet your criteria, but received a 1 seed and finished #1 or #2 in KenPom's pre-tourney rankings.



Team
Post-Tourney
Pre-Tourney
Seed
Overall
KP Rank
NCAA finish


2016 Nova
8
22
2
5
NC


2015 Kentucky
7
7
1
1
FF


2013 Louisville
8
18
1
2
NC


2012 Kentucky
9
8
1
1
NC


2010 Duke
6
9
1
2
NC


2008 Kansas
3
4
1
1
NC


2008 Memphis
8
26
1
3
FF/CG


2007 UNC
7
5
1
1
E8


2005 UNC
5
9
1
2
NC


2005 Illinois
7
7
1
1
FF/CG


2004 Duke
5
6
1
1
FF


2003 Kentucky
9
12
1
1
E8


2002 Duke
1
3
1
1
S16


2010 Kansas
10
6
1
1
R32


2002 Cincy
10
8
1
2
R32

Acymetric
12-13-2018, 12:23 PM
So, this conversation has got me thinking. I wonder if there are any performance differences for teams with wider variance between offensive and defensive ratings. Example:

Team A:




Overall
Offense
Defense


Team A
22
64
7


Team B
23
26
26


Team C
28
15
73



Specifically, I kind of wonder if higher variance between offense defense might make a team more (or less) vulnerable to upsets. Is it better to be highly specialized on one end but not good on the other or to be more balanced?

Philadukie
12-13-2018, 12:44 PM
So, this conversation has got me thinking. I wonder if there are any performance differences for teams with wider variance between offensive and defensive ratings. Example:

Team A:




Overall
Offense
Defense


Team A
22
64
7


Team B
23
26
26


Team C
28
15
73



Specifically, I kind of wonder if higher variance between offense defense might make a team more (or less) vulnerable to upsets. Is it better to be highly specialized on one end but not good on the other or to be more balanced?

Your narrow question alludes to the main question, I think.

Let's set the rankings aside for a second, and just ask the main question in simple terms: is a team who is more balanced on offense and defense, and, specifically in this case, exceptionally balanced on offense and defense, more likely to succeed in the tournament than a team who's more unbalanced on offense and defense? That sentence is more unwieldy than I had planned, but hopefully the logic of the question still comes through.

uh-oh seems to think that the answer is already self-evidently baked into the overall efficiency ranking (and he might be right!) and there' s not much point in looking into it further; while I'm making the case that there's more nuance to the picture and that exceptional offensive and defensive balance found in combining and comparing the AdjO and AdjD rankings has some predictive value that isn't found in the overall ranking.

House_P is unsure and frames the question (and the test of the hypothesis) pretty well above.

I'm not prepared to take the argument (or the work of figuring it out) any further. Real work calls! But hopefully this has spurred a conversation that will yield an answer. I've staked out a position in advance of that answer. I'd like to know if I'm right. ;)

Kedsy
12-13-2018, 03:24 PM
Your narrow question alludes to the main question, I think.

Let's set the rankings aside for a second, and just ask the main question in simple terms: is a team who is more balanced on offense and defense, and, specifically in this case, exceptionally balanced on offense and defense, more likely to succeed in the tournament than a team who's more unbalanced on offense and defense? That sentence is more unwieldy than I had planned, but hopefully the logic of the question still comes through.

uh-oh seems to think that the answer is already self-evidently baked into the overall efficiency ranking (and he might be right!) and there' s not much point in looking into it further; while I'm making the case that there's more nuance to the picture and that exceptional offensive and defensive balance found in combining and comparing the AdjO and AdjD rankings has some predictive value that isn't found in the overall ranking.

House_P is unsure and frames the question (and the test of the hypothesis) pretty well above.

I'm not prepared to take the argument (or the work of figuring it out) any further. Real work calls! But hopefully this has spurred a conversation that will yield an answer. I've staked out a position in advance of that answer. I'd like to know if I'm right. ;)

Using pre-tournament rankings:

Total oRtg+dRtg <= 10
15 teams (since 2002): 4 champs (26.7%); 1 runner-up (6.7%); 4 lost in Final Four (60.0% Final Four teams); 3 Elite Eight (20.0%); 1 Sweet 16 (6.7%); 2 R32 (13.3%)

Total oRtg+dRtg > 10 and <= 20
35 teams (since 2002): 3 champs (8.6%); 6 runners-up (17.1%); 2 lost in Final Four (31.4% Final Four teams); 11 Elite Eight (31.4%); 7 Sweet 16 (20.0%); 5 R32 (14.3%); 1 R64 (2.9%)

Total oRtg+dRtg <= 20
50 teams (since 2002, including the 15 teams in the first category): 7 champs (14%); 7 runners-up (14%); 6 lost in Final Four (40% F4 teams); 14 E8 (28%); 8 S16 (16%); 7 R32 (14%); 1 R64 (2%)

1-seeds that were #1 or #2 in KenPom and total oRtg+dRtg > 10
15 teams (since 2002): 2 champs (13.3%); 3 runners-up (20.0%); 0 lost in F4 (33.3% F4 teams); 5 E8 (33.3%); 3 S16 (20.0%); 1 R32 (6.7%); 1 R64 (6.7%)

Pomeroy top 5
85 teams (since 2002): 12 champs (14.1%); 8 runners-up (9.4%); 10 lost in F4 (35.3% F4 teams); 18 E8 (21.2%); 17 S16 (20.0%); 15 R32 (17.6%); 5 R64 (5.9%)

Pomeroy top 2
34 teams (since 2002): 6 champs (17.6%); 4 runners-up (11.8%); 4 lost in F4 (41.2% F4 teams); 9 E8 (26.5%); 6 S16 (17.6%); 3 R32 (8.8%); 2 R64 (5.9%)

All 1-seeds
68 teams (since 2002): 11 champs (16.2%); 6 runners-up (8.8%); 9 lost in F4 (38.2% F4 teams); 20 E8 (29.4%); 11 S16 (16.2%); 10 R32 (14.7%); 1 R64 (1.5%)



So, it looks like picking a total KenPom rank <= 20 OR KenPom top 5 OR 1-seed have all yielded approximately the same results in the NCAA tourney since 2002.

Total KenPom rank <= 10, however, appears better, although there might not be enough datapoints to be significant.

Partially answering House P's question, there were exactly the same number of 1-seeds that were #1 or #2 in pre-tourney KenPom and did not meet PhilaDukie's criteria, and that group had fewer champs and fewer Final Four teams. Again, though, it's only 15 datapoints.

Philadukie
12-13-2018, 03:43 PM
Using pre-tournament rankings:

Total oRtg+dRtg <= 10
15 teams (since 2002): 4 champs (26.7%); 1 runner-up (6.7%); 4 lost in Final Four (60.0% Final Four teams); 3 Elite Eight (20.0%); 1 Sweet 16 (6.7%); 2 R32 (13.3%)

Total oRtg+dRtg > 10 and <= 20
35 teams (since 2002): 3 champs (8.6%); 6 runners-up (17.1%); 2 lost in Final Four (31.4% Final Four teams); 11 Elite Eight (31.4%); 7 Sweet 16 (20.0%); 5 R32 (14.3%); 1 R64 (2.9%)

Total oRtg+dRtg <= 20
50 teams (since 2002, including the 15 teams in the first category): 7 champs (14%); 7 runners-up (14%); 6 lost in Final Four (40% F4 teams); 14 E8 (28%); 8 S16 (16%); 7 R32 (14%); 1 R64 (2%)

1-seeds that were #1 or #2 in KenPom and total oRtg+dRtg > 10
15 teams (since 2002): 2 champs (13.3%); 3 runners-up (20.0%); 0 lost in F4 (33.3% F4 teams); 5 E8 (33.3%); 3 S16 (20.0%); 1 R32 (6.7%); 1 R64 (6.7%)

Pomeroy top 5
85 teams (since 2002): 12 champs (14.1%); 8 runners-up (9.4%); 10 lost in F4 (35.3% F4 teams); 18 E8 (21.2%); 17 S16 (20.0%); 15 R32 (17.6%); 5 R64 (5.9%)

Pomeroy top 2
34 teams (since 2002): 6 champs (17.6%); 4 runners-up (11.8%); 4 lost in F4 (41.2% F4 teams); 9 E8 (26.5%); 6 S16 (17.6%); 3 R32 (8.8%); 2 R64 (5.9%)

All 1-seeds
68 teams (since 2002): 11 champs (16.2%); 6 runners-up (8.8%); 9 lost in F4 (38.2% F4 teams); 20 E8 (29.4%); 11 S16 (16.2%); 10 R32 (14.7%); 1 R64 (1.5%)



So, it looks like picking a total KenPom rank <= 20 OR KenPom top 5 OR 1-seed have all yielded approximately the same results in the NCAA tourney since 2002.

Total KenPom rank <= 10, however, appears better, although there might not be enough datapoints to be significant.

Partially answering House P's question, there were exactly the same number of 1-seeds that were #1 or #2 in pre-tourney KenPom and did not meet PhilaDukie's criteria, and that group had fewer champs and fewer Final Four teams. Again, though, it's only 15 datapoints.

Thanks for doing that work! So if at the end of the regular season, Duke is #1 in KenPom, a #1 seed, and has a combined oRtg and dRtg rank <= 10, you would put about equal predictive weight on each of them?

How about coming at it from the second question? Let's hypothetically say Nevada pulls off a #1 seed but stays at it's current overall KenPom rank of #5 and it's current combined oRtg and dRtg of 35 right before the tournament. Would you discount the predicative value of their 1 seed to win it all due to a combined oRtg and dRtg of greater than 20?

(Sorry, feel free to punt on the questions, I don't mean to make you do more work, I'm just genuinely interested).

uh_no
12-13-2018, 04:00 PM
does anyone happen to have historical data on average spread in games played between two teams of arbitrary seeds?

It should be simple to figure out predictive value:

-take the spread calculated by KP based on pre-tournament rankings
-take the average historical spread in games played between teams of the given seeds
-calculate a regression of spread based on difference in cumulative kenpom rankings

Train against a bunch of test tournaments (say year <=2017) and then evaluate against 2018. See which one produces a fit closest to 1-1

I'd be happy to do it if someone had a relatively sanitized tournament game scores stream, and similarly sanitized pre-tournament data stream....but scraping HTML and CSV isn't in the books right now :)

Acymetric
12-13-2018, 04:10 PM
does anyone happen to have historical data on average spread in games played between two teams of arbitrary seeds?

It should be simple to figure out predictive value:

-take the spread calculated by KP based on pre-tournament rankings
-take the average historical spread in games played between teams of the given seeds
-calculate a regression of spread based on difference in cumulative kenpom rankings

Train against a bunch of test tournaments (say year <=2017) and then evaluate against 2018. See which one produces a fit closest to 1-1

I'd be happy to do it if someone had a relatively sanitized tournament game scores stream, and similarly sanitized pre-tournament data stream...but scraping HTML and CSV isn't in the books right now :)

I think to really determine the predictive value you need more than one set to evaluate against as well. Train against odd number years, evaluate against even going back <arbitrary years>?

Kedsy
12-13-2018, 04:10 PM
does anyone happen to have historical data on average spread in games played between two teams of arbitrary seeds?

It should be simple to figure out predictive value:

-take the spread calculated by KP based on pre-tournament rankings
-take the average historical spread in games played between teams of the given seeds
-calculate a regression of spread based on difference in cumulative kenpom rankings

Train against a bunch of test tournaments (say year <=2017) and then evaluate against 2018. See which one produces a fit closest to 1-1

I'd be happy to do it if someone had a relatively sanitized tournament game scores stream, and similarly sanitized pre-tournament data stream...but scraping HTML and CSV isn't in the books right now :)

I have pre-tournament KenPom data (which I think I got from House P), from 2002 to 2017 (I have the 2018 pre-tourney data as well, but it's not in the same spreadsheet). I started putting together a spreadsheet of tournament games, but I never finished it (didn't get too far with it at all, actually). If anyone else has one, I'd love to get my virtual hands on it.

Kedsy
12-13-2018, 04:24 PM
Thanks for doing that work! So if at the end of the regular season, Duke is #1 in KenPom, a #1 seed, and has a combined oRtg and dRtg rank <= 10, you would put about equal predictive weight on each of them?

Well, like I said, oRtg+dRtg <= 10 has had better results, but there are so few datapoints it's hard to say if the better performance is real or not. oRtg+dRtg<=20 results were about the same as #1 seed and top 5 (or top 2) KenPom. I can't say for certain, but it didn't appear as if combining these factors would give us significantly better results, at least not without dragging the number of datapoints to an unreliably low level.


How about coming at it from the second question? Let's hypothetically say Nevada pulls off a #1 seed but stays at it's current overall KenPom rank of #5 and it's current combined oRtg and dRtg of 35 right before the tournament. Would you discount the predicative value of their 1 seed to win it all due to a combined oRtg and dRtg of greater than 20?

I absolutely would not discount the predictive value of a #1 seed. The numbers I calculated above were looking at each thing independently. Historically, independent of rating, 16% of 1-seeds have won the tournament. Independent of seed, 14% of KenPom top 5 teams have won the tournament. Both those numbers are independent of combined oRtg+dRtg, so I wouldn't discount either of them based on that factor.

House P
12-13-2018, 04:45 PM
does anyone happen to have historical data on average spread in games played between two teams of arbitrary seeds?

It should be simple to figure out predictive value:

-take the spread calculated by KP based on pre-tournament rankings
-take the average historical spread in games played between teams of the given seeds
-calculate a regression of spread based on difference in cumulative kenpom rankings

Train against a bunch of test tournaments (say year <=2017) and then evaluate against 2018. See which one produces a fit closest to 1-1

I'd be happy to do it if someone had a relatively sanitized tournament game scores stream, and similarly sanitized pre-tournament data stream...but scraping HTML and CSV isn't in the books right now :)

This isn't exactly what you requested, but it may (or may not) be helpful. As Kedsy mentioned, I have a spreadsheet with all the pre-tourney KenPom ratings between 2002 and 2018. This includes the seed each team received, but does not include tourney results.

Anyway, here is a table with the average Adj EM by seed.





Seed
Mean
AdjEM
Std Dev


1
28.52
3.2854


2
24.96
2.7623


3
22.31
2.8097


4
21.65
3.2293


5
20.07
2.8274


6
18.52
2.9746


7
17.37
2.9409


8
16.86
2.7858


9
15.74
2.6023


10
15.79
3.2142


11
14.58
3.0806


12
13.10
3.3644


13
9.57
3.4316


14
7.39
3.688


15
2.27
3.2704


16
-3.44
4.9216

Philadukie
12-13-2018, 04:49 PM
Well, like I said, oRtg+dRtg <= 10 has had better results, but there are so few datapoints it's hard to say if the better performance is real or not. oRtg+dRtg<=20 results were about the same as #1 seed and top 5 (or top 2) KenPom. I can't say for certain, but it didn't appear as if combining these factors would give us significantly better results, at least not without dragging the number of datapoints to an unreliably low level.



I absolutely would not discount the predictive value of a #1 seed. The numbers I calculated above were looking at each thing independently. Historically, independent of rating, 16% of 1-seeds have won the tournament. Independent of seed, 14% of KenPom top 5 teams have won the tournament. Both those numbers are independent of combined oRtg+dRtg, so I wouldn't discount either of them based on that factor.

Thanks. Something just doesn't sit intuitively well about saying, in a hypothetical scenario carried forward from today, that Duke, with a combined oRtg+dRtg of 7, and Nevada, with a combined oRtg+dRtg of 35, would have equal chance of winning it all as both #1 seeds and both top 5 KenPom overall rankings. But I guess that's just a discomfort I'll have to live with, because, as you note, the things are independent of each other, regardless of my feelings about them.

Kedsy
12-13-2018, 05:35 PM
Thanks. Something just doesn't sit intuitively well about saying, in a hypothetical scenario carried forward from today, that Duke, with a combined oRtg+dRtg of 7, and Nevada, with a combined oRtg+dRtg of 35, would have equal chance of winning it all as both #1 seeds and both top 5 KenPom overall rankings. But I guess that's just a discomfort I'll have to live with, because, as you note, the things are independent of each other, regardless of my feelings about them.

It's the nature of the one-and-done tournament. #1 seeds are in the best position to get to the Final Four. The best teams are the best teams, and thus have the best chance to win. But game-to-game variability, differing quality in opponent (due to odd seeding or, more frequently, earlier upsets), and just plain luck play a much larger role in who wins and who loses than many people feel or believe. The best team rarely wins the NCAA tournament. Fact is, KenPom's pre-tournament #1 team has only won the national championship twice in 17 years (11.8%).

House P
12-19-2018, 11:12 AM
Thanks to Troublemaker's guidance on finding preseason KenPom ratings, I made a very rough estimate of how much his current ratings are still influenced by his preseason ratings. A short description of my methodology is below, but here is the takeaway.

It appears that approximately 50% of KenPom's current ratings are still based on his preseason ratings.



I did some additional digging into KenPom's blog posts and found a different (better?) way to estimate how much preseason weighting is currently factored into KenPom's ratings. Based on this approach (methodology below), I found the following.

It appears that approximately 25-30% of KenPom's ratings on Dec 26, 2012 were based on his preseason ratings.

So, as Uh_No pointed out, my previous post likely overestimated the current influence of preseason ratings. However, given all of Duke's blowouts (which aren't weighted as full games in KenPom's methodology), I would still estimate that roughly a quarter to a third of Duke's current KenPom rating is still based on pre-season ratings. So, when someone such as Kedsy points out that Duke's KenPom defensive rating is now ahead of UVA's, it should probably be noted that Duke would be even further ahead if preseason expectations (UVA = 86.5, Duke = 89.6) weren't currently being factored in.

That being said, while removing the influence of pre-season ratings likely provides a better description of a team's performance to date, the use of preseason ratings likely results in a better prediction of a team's performance going forward.




METHODOLOGY

In this post from Dec 26, 2012 (https://kenpom.com/blog/preseason-ratings-why-weight/), Ken provides tables which list 20 teams with the most variation between their rankings with and without preseason weighting through Dec 26, 2011. While his tables don't include each team's Adjusted Efficiency Margin (AdjEM) without a without preseason weighting, he does list where each team would be ranked with and without preseason weighting. Therefore, each team's AdjEM rating with and without preseason weighting can be estimated by looking at the AdjEM associated with a particular spot in the actual rankings.

For example, KenPom mentions that on Dec 26, 2011 Maryland was ranked 166, but would have been ranked at 236 if Maryland's preseason ranking of 47 wasn't considered. Using KenPom's rating archives, we know that the #166 team in the Dec 26 rankings had an AdjEM of 0.07 while the #236 team had a AdjEM of -6.25. So, when you consider that Maryland had a preseason AdjEM of 13.13, you can estimate that preseason ratings contributed ~33% towards Maryland's rating on Dec 26, 10 games into their season.

I did this for every team with a difference of at least 20 places between their rank with and without preseason weighting and ended up with an average preseason weighting of 27% (standard deviation = 6.6%, range = 17% - 41%).

I should note that the blog post was from 2012 and Ken has tweaked his methodology since then. So he may have also changed his preseason weighting component.

JayZee
12-19-2018, 12:04 PM
And as pointed out by others, relative AdjEM each year is an important measure and right now we are more than 5 pts better than the #2 team. That is a huge spread.

Duke 2002 and 2004 had 4 pt spreads

Ohio State 2011 had a 5 point spread (over #2 duke...)

Other than that there were some 3 point spreads, but usually much closer at the top

wtrimble99
12-19-2018, 03:18 PM
I did some additional digging into KenPom's blog posts and found a different (better?) way to estimate how much preseason weighting is currently factored into KenPom's ratings. Based on this approach (methodology below), I found the following.

It appears that approximately 25-30% of KenPom's ratings on Dec 26, 2012 were based on his preseason ratings.

So, as Uh_No pointed out, my previous post likely overestimated the current influence of preseason ratings. However, given all of Duke's blowouts (which aren't weighted as full games in KenPom's methodology), I would still estimate that roughly a quarter to a third of Duke's current KenPom rating is still based on pre-season ratings. So, when someone such as Kedsy points out that Duke's KenPom defensive rating is now ahead of UVA's, it should probably be noted that Duke would be even further ahead if preseason expectations (UVA = 86.5, Duke = 89.6) weren't currently being factored in.

That being said, while removing the influence of pre-season ratings likely provides a better description of a team's performance to date, the use of preseason ratings likely results in a better prediction of a team's performance going forward.

Hey everybody. First post here but longtime lurker.

Your effort to find surrogate KenPom rankings that didn't include a preseason weighting reminded me of these Barttorvik.com rankings (http://barttorvik.com/trank.php?year=2019&sort=&lastx=0&hteam=&conlimit=All&state=All&begin=20181102&end=20190501&top=0&quad=4&venue=All&type=All&mingames=0#) that I've been using for the past few seasons to find reliable non-conference rankings without the preseason component. It turns out that if you set the start date filter to 11/2 of 11/1, it will eliminate all of his preseason component in the rankings.

Obviously, T-Rank isn't a perfect substitute for KenPom, mainly because the game weightings are slightly different and KenPom caps the margin of victory significantly earlier. However, I've found BartTorvik to be a great free alternative to KenPom, especially for those who prefer the old pythagorean ranking method.

What's interesting is how well the rankings match with your initial estimation, where VT, San Francisco, Utah State, and Nebraska have outperformed all expectations, while Kansas hasn't been as sharp as expected, and UK and Nova barely rank as bubble teams.

JayZee
12-21-2018, 04:06 PM
Hey everybody. First post here but longtime lurker.

Your effort to find surrogate KenPom rankings that didn't include a preseason weighting reminded me of these Barttorvik.com rankings (http://barttorvik.com/trank.php?year=2019&sort=&lastx=0&hteam=&conlimit=All&state=All&begin=20181102&end=20190501&top=0&quad=4&venue=All&type=All&mingames=0#) that I've been using for the past few seasons to find reliable non-conference rankings without the preseason component. It turns out that if you set the start date filter to 11/2 of 11/1, it will eliminate all of his preseason component in the rankings.

Obviously, T-Rank isn't a perfect substitute for KenPom, mainly because the game weightings are slightly different and KenPom caps the margin of victory significantly earlier. However, I've found BartTorvik to be a great free alternative to KenPom, especially for those who prefer the old pythagorean ranking method.

What's interesting is how well the rankings match with your initial estimation, where VT, San Francisco, Utah State, and Nebraska have outperformed all expectations, while Kansas hasn't been as sharp as expected, and UK and Nova barely rank as bubble teams.

Thanks for posting.

What is super cool about that website is that you can chart offensive/defensive efficiency across a season. For instance, you can see how 2015, about mid ACC season, went from pretty bad and kept improving all the way through the NC game.

It's about the journey.

timmy c
01-04-2019, 12:25 AM
Zion Williamson moved into the lead tonight in kenpom's player of the year race. Former leader, Ethan Happ, put up a pedestrian numbers in the loss to Minnesota. This allowed Zion to move ahead of him. Interesting to note that despite some of the consternation about RJ's performances, he is now rated number two in the race. I don't believe I have ever seen two teammates in the top two spots. Will these two be able to hold on to win ken pomroy's title?

BandAlum83
01-04-2019, 12:28 AM
Zion Williamson moved into the lead tonight in kenpom's player of the year race. Former leader, Ethan Happ, put up a pedestrian numbers in the loss to Minnesota. This allowed Zion to move ahead of him. Interesting to note that despite some of the consternation about RJ's performances, he is now rated number two in the race. I don't believe I have ever seen two teammates in the top two spots. Will these two be able to hold on to win ken pomroy's title?

Are individual stats only available to subscriber's? I can't find them.

If subscription only, it would be great if you could keep us updated periodically.

Troublemaker
01-04-2019, 07:35 AM
Are individual stats only available to subscriber's? I can't find them.

If subscription only, it would be great if you could keep us updated periodically.

Yes, the KPOY page appears to be subscribers-only. I think you can count on occasional updates, as I think I remember KPOY getting brought up when POY discussions occurred in past seasons.

CDu
01-04-2019, 09:17 AM
Thanks. Something just doesn't sit intuitively well about saying, in a hypothetical scenario carried forward from today, that Duke, with a combined oRtg+dRtg of 7, and Nevada, with a combined oRtg+dRtg of 35, would have equal chance of winning it all as both #1 seeds and both top 5 KenPom overall rankings. But I guess that's just a discomfort I'll have to live with, because, as you note, the things are independent of each other, regardless of my feelings about them.

I wouldn't say that Duke and Nevada have the same chance of winning it all, and the data Kedsy provided don't say that either. Duke, for example, checks another box that Nevada doesn't (KenPom top-10 in both offense and defense). That further separates them from the Wolf Pack. And to Kedsy's question, I do believe the difference in results between the "top-10 in O and D" and the other groups is a significant difference. We could take it a step further and look at teams that were KenPom top-5 in both offense and defense, but that certainly won't have enough data points to be significant. But I'd imagine that teams that are top-5 in both would fare even better than teams that are top-10 in both if we had a sufficient sample of data points (which we clearly don't if only 15 teams met the top-10 threshold).

Furthermore, nothing in the idea of those groupings says "all teams with those characteristics have the same chance of winning". It's just that, as a collective, teams in those groups have historically done this well. Some teams in that group would be expected to do better, while others in the group drag the average down. So even if Duke wasn't top-10 in both defense and offense, it wouldn't be appropriate to conclude they have the same chance of winning based on those data. Duke, by virtue of being the #1 team, would have the better chance of winning if they were to play the tournament today.

BoiseDevil
01-04-2019, 12:10 PM
Zion Williamson moved into the lead tonight in kenpom's player of the year race. Former leader, Ethan Happ, put up a pedestrian numbers in the loss to Minnesota. This allowed Zion to move ahead of him. Interesting to note that despite some of the consternation about RJ's performances, he is now rated number two in the race. I don't believe I have ever seen two teammates in the top two spots. Will these two be able to hold on to win ken pomroy's title?

Perfect example of being a spoiled Duke fan...

After reading Z and RJ were 1 and 2... my first thought wasn’t “I hope they can maintain it” but rather “I wonder if Tre or Cam can climb up to #3 if their shots start to fall”

We are living in the Duke gilded age and I love it!

One of the smart posters here commented during the JJ/Shelden regular season to “enjoy it cause it can’t last forever”... That was nearly 13 years ago and what a ride it’s been!

I’m enjoying this season as much as any!

Go Duke!

hallcity
01-04-2019, 01:18 PM
Perfect example of being a spoiled Duke fan...

After reading Z and RJ were 1 and 2... my first thought wasn’t “I hope they can maintain it” but rather “I wonder if Tre or Cam can climb up to #3 if their shots start to fall”

We are living in the Duke gilded age and I love it!

One of the smart posters here commented during the JJ/Shelden regular season to “enjoy it cause it can’t last forever”... That was nearly 13 years ago and what a ride it’s been!

I’m enjoying this season as much as any!

Go Duke!

I keep thinking of the lyrics of an old song — “These are the good old days.” Enjoy them while they last for they cannot last forever.

BandAlum83
01-04-2019, 01:22 PM
Perfect example of being a spoiled Duke fan...

After reading Z and RJ were 1 and 2... my first thought wasn’t “I hope they can maintain it” but rather “I wonder if Tre or Cam can climb up to #3 if their shots start to fall”

We are living in the Duke gilded age and I love it!

One of the smart posters here commented during the JJ/Shelden regular season to “enjoy it cause it can’t last forever”... That was nearly 13 years ago and what a ride it’s been!

I’m enjoying this season as much as any!

Go Duke!

1986 - 1994 were pretty darn gilded as well. 7 final fours, 2 National Championships, 3 ACCT championships, 3 Naismith award winners, 3 National Defensive POY awards, 5 lottery picks.

I guess We have seen a couple of gilded ages for Duke basketball.

91_92_01_10_15
01-04-2019, 02:56 PM
Are individual stats only available to subscriber's? I can't find them.

If subscription only, it would be great if you could keep us updated periodically.

If you are a subscriber, here is the link:

https://kenpom.com/kpoy.php

HereBeforeCoachK
01-04-2019, 03:58 PM
1986 - 1994 were pretty darn gilded as well. 7 final fours, 2 National Championships, 3 ACCT championships, 3 Naismith award winners, 3 National Defensive POY awards, 5 lottery picks.l.

sooooooooo close to 3 natty in 4 years....woulda been so sweet...on tee shirts, coffee mugs, etc

91 DUKE
92 DUKE
93 FLUKE
94 DUKE