Counting down the unbeatens

[Olympic Fan]

Nobody has made it through the season unbeaten since Indiana in 1976. A lot of Kentucky fans -- and some national commentators -- think Kentucky can do it this year. We'll see.

I've always found it fun to count down the unbeaten. As of today (Dec. 8) we're down to 12 unbeaten teams. It's an eclectic mix, ranging from the true powers (Kentucky, Duke, Arizona ...) to some really funky nobodies.

One of them is a first-year Division 1 team called Incarnate Word of San Antonio, Tex. They belong to the Southland Conference. Their presence on this list is really a fluke. Because of their transition to Division 1, they play a strange schedule. They are 5-0 at the moment, but four of the wins have come against non-Division 1 teams. Their one legit win is over Princeton. Now they start a string of seven straight road games in which they WILL lose. Probably tonight when they play at UTEP ... if not then, Wednesday night at Nebraska.

They are the reason I'm posting now, instead of waiting a few days. I wanted to mention them before they slip out of the ranks of the unbeaten.

Three more unbeaten could go down Tuesday night (Dec. 9). Seton Hall is at No. 11 Wichita State, while Louisville meets Indiana in Madison Square Garden. Villanova takes on Illinois also in MSG. I think the Hall goes down, while the Ville and Nova remains unbeaten, but none of the three games is a lock.

The eight other unbeaten and their next real chance to lose:
Duke (Dec. 18 vs. UConn in New York)
Kentucky (UNC on Dec. 13; UCLA on Dec. 20; at Louisville on Dec. 27)
Virginia (Jan. 3 at Miami)
TCU (Jan. 3 vs. West Virginia; Jan. 7 at Kansas State)
Northern Iowa (Dec. 13 at VCU)
Colorado State (Dec. 10 at Colorado)
Arizona (Dec. 19 at UTEP; Dec. 23 at UNLV)
Washington (Dec. 20 vs. Oklahoma)

If Louisville and Villanova survive tomorrow night as I expect, their next chances to lose come:
Louisville (Dec. 27 vs. Kentucky)
Villanova (Dec. 20 vs. Syracuse)

I think at least a half dozen of these teams (including Duke) make it to January unbeaten, but I'll be really impressed if anybody -- including Kentucky -- makes it to February with a perfect record).

I will say this -- with the weakness of the SEC, if Kentucky gets past Louisville unbeaten on the 27thy, they could go a long, long way without a loss.

PS Holy cow! I just realized that the Kentucky-Louisville game on Dec. 27 is the exact same start time as the Duke-Arizona State Sun Bowl! Thank god for DVRs.

[Wander]

Does TCU have the single worst non-conference schedule in Division 1? Right now, it's ranked 340th (out of 351) on kenpom... but that doesn't include five remaining games, all against terrible teams. The BEST remaining opponent is 271 McNeese State. The other remaining teams are ranked 295, 314, 325, and last place in all D1, 351. Those games are all at home. Wow, TCU.

[pfrduke]

One of them is a first-year Division 1 team called Incarnate Word of San Antonio, Tex. They belong to the Southland Conference. Their presence on this list is really a fluke. Because of their transition to Division 1, they play a strange schedule. They are 5-0 at the moment, but four of the wins have come against non-Division 1 teams. Their one legit win is over Princeton. Now they start a string of seven straight road games in which they WILL lose. Probably tonight when they play at UTEP ... if not then, Wednesday night at Nebraska.

This is actually year 2 for Incarnate Word. They got off to a similarly (although not identically) good start last season, turning in a 12-1 non-conference record. However, the 12 were all against non-DI teams and the only DI opponent they faced (North Texas) was responsible for the loss. This year's non-conference schedule, with "only" 4 games against non-DI opponents, is a marked improvement over last year's.

[ACCBBallFan]

Of course conference play with be battle of unbeatens and dwindle this list

Duke plays UVA J31 and Lville J17 and they play one another on F07 and M07.

Zona faces Wash F13

If UK gets by the 2 tough ACC teams, UNC and L-ville, Florida will be better when Alex Murphy is eligible and palyers return from injury

Higher ranked once beaten Wich St plays N Iowa J31 and F28/

Higher ranked KU, Texas, Baylor among others in B12 play TCU twice.

Higher ramked SD St and Wyming play home and homes with Coloradi St, etc.

[arnie]

Does TCU have the single worst non-conference schedule in Division 1? Right now, it's ranked 340th (out of 351) on kenpom... but that doesn't include five remaining games, all against terrible teams. The BEST remaining opponent is 271 McNeese State. The other remaining teams are ranked 295, 314, 325, and last place in all D1, 351. Those games are all at home. Wow, TCU.

Looks like the Big 10 composite OOC football wins

[Olympic Fan]

Down to 11 unbeaten ... UTEP knocks of Incarnate Word as expected.

[Olympic Fan]

As expected, Seton Hall falls to Wichita State, but Villanova and Louisville win ...

That leaves 10 unbeaten in Division 1

[subzero02]

Indiana made it interesting at the start of the 2nd half but they couldn't keep pace with the Cardinals and their uncontested dunks.

[Wander]

Well, maybe Incarnate Word wasn't so bad after all!

Or more likely, I was wrong about Nebraska, and they suck. Had the ball with 6 seconds left up by 1, has no idea how to throw the ball in bounds.

[Olympic Fan]

Northern Iowa fell in double OT to BCU Saturday.

That leaves nine unbeaten.

[Henderson]

Northern Iowa fell in double OT to VCU Saturday.

That leaves nine unbeaten.

There you go.

[Wander]

Can I complain some more about TCU? Their schedule is now down to 349th out of 351 teams, and they still haven't played 297, 319, and 351, all at home. This has to be historic in some way.

Slightly more on topic, Washington should watch out for Eastern Washington today, which has already won at Indiana this season.

[AncientPsychicT]

Can I complain some more about TCU? Their schedule is now down to 349th out of 351 teams, and they still haven't played 297, 319, and 351, all at home. This has to be historic in some way.

Slightly more on topic, Washington should watch out for Eastern Washington today, which has already won at Indiana this season.

I don't think Washington should be too worried. It seems like everyone and their mother is winning games at B1G teams this year. Well, except Syracuse, that is.

[pfrduke]

Can I complain some more about TCU? Their schedule is now down to 349th out of 351 teams, and they still haven't played 297, 319, and 351, all at home. This has to be historic in some way.

Slightly more on topic, Washington should watch out for Eastern Washington today, which has already won at Indiana this season.

Eastern's a legitimately good team - top 100 in KenPom and an excellent offense, scoring over 80 in all but 2 of their games. Should be a clash of strengths - the Huskies' FG defense has been excellent and the Eagles have been very effective at putting the ball in the hoop. I think the Huskies can do enough on offense to win, but I would not at all be surprised to see a close game.

[CharlestonDave]

This is just one person's opinion, without any substantiation.

I think we will lose at Louisville, at Virginia, split with UNC, and possibly lose at Notre Dame.

We will have a tough game against Miami at home, a tough game at the Garden against St. John's and a surprisingly tough game at the Carrier Dome against Syracuse.

Just an opinion that's all.

[Mtn.Devil.91.92.01.10.15]

This is just one person's opinion, without any substantiation.

I think we will lose at Louisville, at Virginia, split with UNC, and possibly lose at Notre Dame.

We will have a tough game against Miami at home, a tough game at the Garden against St. John's and a surprisingly tough game at the Carrier Dome against Syracuse.

Just an opinion that's all.

Two things:

1) I don't believe the intention of this thread is to imply that Duke will run the table. Rather, I understand it as a way to keep some interesting perspective on the rest of the country; particularly when so many media yammering heads are calling for UK as the best team ever.

However...

2) If your predictions above involve a Duke team that is currently #2 in the nation with a perfect record losing half their games in January and February, it's possible you are on the wrong message board. I believe everyone here would be devestated to be just over .500 in conference play, given how our season has started.

*** edit ***
I am not saying that your predictions are necessarily wrong, but since you admit they are unsubstantiated and since that outcome would be extremely disappointing to the members of this board, it smells like trolling. If it isn't, I apologize for the snark and admit you are welcome to your own opinion. In light of such, I predict Kentucky will beat UNC in the NIT final.

[AncientPsychicT]

a surprisingly tough game at the Carrier Dome against Syracuse.

Why would you call this game surprisingly tough? Syracuse was supposed to be pretty good this year, but so far they've stunk. If anything, this game should be surprisingly easy compared to preseason expectations.

[JasonEvans]

Can I complain some more about TCU? Their schedule is now down to 349th out of 351 teams, and they still haven't played 297, 319, and 351, all at home. This has to be historic in some way.

Historic? In what way?

You are aware that some team has to have the worst schedule in the land, right? TCU may be 349th, but someone else is #350 and yet another team is #351. In fact, do you know who is #350? Notre Dame.

-Jason "what's really pathetic is that Va Tech is #345, a truly wretched and easy non-conference schedule... and yet they are only 5-4 against that sked. Ouch!" Evans

[Skitzle]

If I'm not mistaken the complaint is that even though they have almost the easiest schedule possible, they are unbeaten and getting recognition for it.

All I can say is patience. It is highly improbably that TCU goes unbeaten. I'd bet on Kentucky before TCU schedule and all.

[Henderson]

If I'm not mistaken the complaint is that even though they have almost the easiest schedule possible, they are unbeaten and getting recognition for it.

All I can say is patience. It is highly improbably that TCU goes unbeaten. I'd bet on Kentucky before TCU schedule and all.

Clemson used to schedule that way too. You'd see them climb in the rankings with a gaudy record, then **pop** and back to your regularly scheduled programming. It might be a way to best prepare for the season; it might not be much preparation at all. I'll leave that to coaches to sort out. But it does get folks talking about your team and school.

[bbosbbos]

He may say whatever he wants while you may predict whatever you want. :p May I predict: (1) Tar Heels will be banned for 10 yrs in college sports? (2) Caripali will retire at the end of the season? :p


I predict Kentucky will beat UNC in the NIT final.

[Wander]

Historic? In what way?

You are aware that some team has to have the worst schedule in the land, right? TCU may be 349th, but someone else is #350 and yet another team is #351. In fact, do you know who is #350? Notre Dame.


Fair enough, but (1) I'm pretty sure TCU will pass Notre Dame in the next few weeks, and (2) there are also absolute values associated with SOS, not just relative. So it might be the case that we can compare across years and find that TCU had the worst SOS of any BCS team in the kenpom era when the non-conference season is over.

Also... TCU is now "ranked" 27th if you count others receiving votes. Ewwwwww.

[CharlestonDave]

Two things:

1) I don't believe the intention of this thread is to imply that Duke will run the table. Rather, I understand it as a way to keep some interesting perspective on the rest of the country; particularly when so many media yammering heads are calling for UK as the best team ever.

However...

2) If your predictions above involve a Duke team that is currently #2 in the nation with a perfect record losing half their games in January and February, it's possible you are on the wrong message board. I believe everyone here would be devestated to be just over .500 in conference play, given how our season has started.

*** edit ***
I am not saying that your predictions are necessarily wrong, but since you admit they are unsubstantiated and since that outcome would be extremely disappointing to the members of this board, it smells like trolling. If it isn't, I apologize for the snark and admit you are welcome to your own opinion. In light of such, I predict Kentucky will beat UNC in the NIT final.

I was not trolling and probably yes the outcome would be disappointing both to me and to members of the Board. In the latest college polls both Louisville and Virginia are ranked in the top 10 and losing to either of them at home would not be devastating.

I would love for Duke to duplicate the Indiana record of years back . What matters for this team is how they do in March in the tournament and as of right now they look like a Final Four team.

[CharlestonDave]

Why would you call this game surprisingly tough? Syracuse was supposed to be pretty good this year, but so far they've stunk. If anything, this game should be surprisingly easy compared to preseason expectations.

You are correct in your assessment of Syracuse . I just think that Boeheim will get his team sky high at the Carrier Dome to play his good friend . Certainly, I hope that I am wrong.

[ACCBBallFan]

The only team Ken Pomeroy gives a relatively high chance of going undefeated prior to league tournaments is UK

13.5% UK but has them losing to Lville 65-64

1.7 % Duke but has them losing to L-ville 72-70

1.4 % UVA but again has them losing to Lville in last regular season game 59-57

1.3 % Zona favored in all remaining games with 52% chance of winning at Utah

1.1 % Nova favored in all remaining gaes with 56% chance of winning at G-town, 58 % vs. St. John's

0.5 % Lville but has them losing first home and home to UVA 60-56

0.02 % Colorado St first loss 76-74 at New MX St

0.0 % Washington first loss 73-69 to Oklahoma on neutral court

0.0% TCU first loss 63-60 at K-State

[mgtr]

So, Louisville will be the spoiler for many teams this season, if KP analysis turns out. I have seen L'ville play parts of two games, they will be a tough out, but so will Duke. UNC - not so much!

[DukeDevil]

So, Louisville will be the spoiler for many teams this season, if KP analysis turns out. I have seen L'ville play parts of two games, they will be a tough out, but so will Duke. UNC - not so much!

If you look at the subgroup analysis for Louisville, it turns out they beat themselves in the Virginia game

[ACCBBallFan]

The only team Ken Pomeroy gives a relatively high chance of going undefeated prior to league tournaments is UK

13.5% UK (29-2 overall, 16-2 SEC) ) but has them losing to Lville 65-64

1.7 % Duke (27-4, 15-3)) but has them losing to L-ville 72-70

1.4 % UVA (26-4, 15-3) but again has them losing to Lville in last regular season game 59-57

1.3 % Zona (27-4, 14-4) favored in all remaining games with 52% chance of winning at Utah

1.1 % Nova (27-4, 14-4) favored in all remaining gaes with 56% chance of winning at G-town, 58 % vs. St. John's

0.5 % Lville (26-5, 14-4) but has them losing first home and home to UVA 60-56

0.02 % Colorado St (24-7, 11-7) first loss 76-74 at New MX St

0.0 % Washington (22-8, 11-7) first loss 73-69 to Oklahoma on neutral court

0.0% TCU (19-12) first loss 63-60 at K-State

I added the projected records to prior post above, and here are some teams that have already lost that are projected by Ken Pom to win their conference:

Wisc (27-4, 15-3)
Zags (28-3, 16-2)
TX (24-7, 12-6)
KU (24-7, 12-6)
Wich St (25-3, 16-2)
Harvard (23-5, 12-2)
Wisc - GB (23-6, 12-4)
S F Austin (25-6, 16-2)
GA State (24-7, 17-3)
San Diego St (24-7, 14-4)
E. Wash (23-8, 14-4)
V C U (23-8, 13-5)
N C Central (22-8, 14-2)
FL Gulf Coast (22-8, 13-3)
UConn (19-11, 13-5)
Coastal Carolina (22-8, 13-5)

[Tripping William]

If you look at the subgroup analysis for Louisville, it turns out they beat themselves in the Virginia game

Was Ademola Okulaja somehow involved?

[Mtn.Devil.91.92.01.10.15]

I was not trolling and probably yes the outcome would be disappointing both to me and to members of the Board. In the latest college polls both Louisville and Virginia are ranked in the top 10 and losing to either of them at home would not be devastating.

I would love for Duke to duplicate the Indiana record of years back . What matters for this team is how they do in March in the tournament and as of right now they look like a Final Four team.

I apologize if I was too harsh. Perhaps it was your choice of words saying "I think" as though you've come to this conclusing through a logical analysis, and you stated clearly that it was without substantiation. Maybe if you phrased it "what if," or said something like "these games will be our challenging matchups."

I just find it difficult to believe that anyone who has watched Duke's first 9 games would predict us to finish with that large a number of losses. Could it happen? Certainly.

[Edouble]

I just find it difficult to believe that anyone who has watched Duke's first 9 games would predict us to finish with that large a number of losses. Could it happen? Certainly.

How many losses do you think we will have?

CharlestonDave predicted 3 or 4 losses: "I think we will lose at Louisville, at Virginia, split with UNC, and possibly lose at Notre Dame," with tough games at Miami, at St. Johns's and at Syracuse. I assume we are winning these tough games, or else he would call those losses too.

I predicted two losses on the season, I think in the pre-Elon thread. Would you agree with that assessment?

[arnie]

How many losses do you think we will have?

CharlestonDave predicted 3 or 4 losses: "I think we will lose at Louisville, at Virginia, split with UNC, and possibly lose at Notre Dame," with tough games at Miami, at St. Johns's and at Syracuse. I assume we are winning these tough games, or else he would call those losses too.

I predicted two losses on the season, I think in the pre-Elon thread. Would you agree with that assessment?

I also thought Charleston Dave's prediction was reasonable. It's a long tough ACC season and coupled with the game at St Johns, I would be happy with only 3-4 losses. That reg season total should give us a 1 seed and keep us out of the dreaded 2-15 matchup.

[OZZIE4DUKE]

We will go 40-0. Kentucky will go 39-1. So it is written, so it shall be. GTHc 9F!

[rsvman]

We will go 40-0. Kentucky will go 39-1. So it is written, so it shall be. GTHc 9F!

Ozzie, too bad it is too late for UNC to 0-40.

They could still lose all their remaining games, though, so there's that. :cool:

[SCMatt33]

Ozzie, too bad it is too late for UNC to 0-40.

They could still lose all their remaining games, though, so there's that. :cool:

They don't need to lose the rest of their games, just lose in the semifinals (T-3rd) of the NIT. They deserve a banner season, too.

[ACCBBallFan]

Looks like each of the unbeatens have a Merry Christmas and enjoy home cooking and/or byes except Zona plays at UNLV.

At least one goes down next Saturday as UK visits L-ville, and another (Colorado St) not favored to win

Date Rk Opponent Result Location

Sun Dec 21 UVA 26 Harvard W, 58-48 89% Home

Mon Dec 22 Color St 208 Charleston Southern W, 77-64 90% Home
Mon Dec 22 Washington 172 Tulane W, 75-61 91% Home
Mon Dec 22 T C U 351 Grambling St. W, 74-48 99% Home

Tue Dec 23 Zona 135 UNLV W, 71-61 84% Away
Tue Dec 23 NOVA 269 NJIT W, 85-59 98% Home
Tue Dec 23 L-Ville 268 Cal St. Northridge W, 82-56 99% Home

Sat Dec 27 L-Ville 1 Kentucky L, 64-63 45% Home
Sat Dec 27 Color St 101 New Mexico St. L, 77-76 45% Away
Sat Dec 27 UK 5 Louisville W, 64-63 55% Away

Sun Dec 28 Washington 148 Stony Brook W, 74-62 89% Home

Mon Dec 29 Duke 100 Toledo W, 87-65 96% Home
Mon Dec 29 T C U 326 Tennessee St. W, 72-53 97% Home

Tue Dec 30 UVA 79 Davidson W, 75-58 95% Home
Tue Dec 30 L-Ville 105 Long Beach St. W, 77-59 95% Home

Wed Dec 31 Color St 56 Boise St. W, 70-67 65% Home
Wed Dec 31 NOVA 44 Butler W, 70-59 87% Home
Wed Dec 31 Duke 78 Wofford W, 73-56 95% Home

[mbwalker]

Looks like each of the unbeatens have a Merry Christmas and enjoy home cooking and/or byes except Zona plays at UNLV.


And Zona loses, 71-67. (And that's to a team that lost to Arizona State by 22 on Dec. 3 and Utah by 13 just three days ago. Just because the experts predicted that Arizona had a 84% chance of winning, they still have to play the games.)

[uh_no]

And Zona loses, 71-67. (And that's to a team that lost to Arizona State by 22 on Dec. 3 and Utah by 13 just three days ago. Just because the experts predicted that Arizona had a 84% chance of winning, they still have to play the games.)

Yeah

I wasn't sold on arizona, and seem vindicated....their only real win was gonzaga, who themselves is overhyped IMO....and that was in overtime.....they have a pretty weak schedule the rest of the way, and may yet end up with a 1 seed.....but I would be surprised for them to make the elite 8.

[Saratoga2]

And Zona loses, 71-67. (And that's to a team that lost to Arizona State by 22 on Dec. 3 and Utah by 13 just three days ago. Just because the experts predicted that Arizona had a 84% chance of winning, they still have to play the games.)

This can be read as AZ had a 16% chance of losing and so they did. Sometimes that matchups and/or style of play can confuse a team who is an obvious favorite. We will have many opportunities to stumble as well. So far we have faced teams with different looks and haven't been tripped up. UCONN was a long, quick and aggressive team. Wisconsin an experienced team who had gone far in the tourney last year and if you missed it, Stanford who took down Texas last night. We seem pretty resilient. If we stay healthy we will probably have a low number of losses.

Yes Louisville will be a big challenge and UVA has been hard to solve this year. Yes, there are other teams that can give us trouble. We may well be favored in each of our games unless Louisville can beat Kentucky Saturday. If they do, they will become a favorite over Duke. It would be great to get to the tournament as a #1 seed.

[Kedsy]

This can be read as AZ had a 16% chance of losing and so they did.

Coincidentally, 16% was the same Pomeroy-predicted chance of Duke losing to Mercer last season. And they did also, but it was still pretty surprising.

[OZZIE4DUKE]

Coincidentally, 16% was the same Pomeroy-predicted chance of Duke losing to Mercer last season. And they did also, but it was still pretty surprising.
There is a reason why they play the games on the field and not on paper! LGD!

[AncientPsychicT]

There is a reason why they play the games on the field and not on paper! LGD!

I thought they played them on a court... :confused:

[OldPhiKap]

I thought they played them on a court... :confused:

Ozzie is channeling our football game against the Sun Devils. Basketball begins after that. Go Duke!

[tbyers11]

And then there were 6…

Louisville, as most everyone is aware, lost to Kentucky yesterday.

However, this evening the Washington Huskies lost to Stony Brook (#128 in KPom before the game) 62-57 at home in Seattle. Didn't see the game, but UW appeared to have a comfortable double-digitish lead until the last 5 minutes. They proceeded to go scoreless the last 3:40 of the game and lost by 5.

Washington is not an elite team but had quality wins over San Diego St (home) and Oklahoma (neutral). Stony Brook is not awful but not the type of team you would expect the Huskies to lose their first game against.

----------
Games this week
----------
(#1 KPom) Kentucky - doesn't play until Jan 6th against Ole Miss
(#2) Virginia - 12/30 vs Davidson, 1/3 @ Miami; Miami could be tricky but likely make it through the week
(#3) Duke - 12/29 vs Toledo, 12/31 vs Wofford, 1/3 vs Boston College; KPom actually says that BC is the easiest game of the week (98% vs 96% in the 2 non-con games). Not likely to lose
(#5) Villanova - 12/31 vs Butler, 1/3 @ Seton Hall; KPom only has Nova as 66% favorite against Seton Hall. Nova escaped against Cuse last week. Should be interesting to see how they fare this week
(#47) TCU - 12/29 vs Tenn St, 1/3 vs West Virginia; TCU has the WORST schedule strength (351st) of ALL teams to date. Although KPom has them as 1 pt favorite vs WVa I would be surprised if they win this game
(#57) Colorado St - 12/31 vs Boise St, 1/3 @ New Mexico; 2 conference games this week for CSU. KPom has them as slight underdogs @ UNM in the Pit. I'd guess CSU drops one of these 2 games.

[Wander]

Historic? In what way?

You are aware that some team has to have the worst schedule in the land, right? TCU may be 349th, but someone else is #350 and yet another team is #351. In fact, do you know who is #350? Notre Dame.

-Jason "what's really pathetic is that Va Tech is #345, a truly wretched and easy non-conference schedule... and yet they are only 5-4 against that sked. Ouch!" Evans

TCU's nonconference season is over and they now sit in dead last at 351 and by a decent margin. Their score would have been dead last in 9 of the past 10 seasons. Can I go back to making fun of them now?

PS Your Virginia Tech note now applies to the team at 350.

[Edouble]

Coincidentally, 16% was the same Pomeroy-predicted chance of Duke losing to Mercer last season. And they did also, but it was still pretty surprising.

Paging Duke95...

[Chillduck]

Seton Hall 66 Villanova 61 in OT. And then there were five!

[Duke95]

Paging Duke95...

I remember a few comments here noting that Mercer was very dangerous. Seems somebody's "eye" test was working. ;)

That loss was just painful.

[Wander]

Bye bye, TCU. I bet they need to go 11-7 in the Big 12 to make the tournament.

Colorado State is only favored by a point tonight.

[Duke3517]

If Duke is equal or less than 3 losses would that be considered as the best regular season ever?

[Mabdul Doobakus]

UVA going to OT against Miami. Gritty performance by Miami. I think they were down 38-20 at or maybe shortly after halftime, but from that point they quickly brought the game within 10 and then kept grinding. Virginia hit two big threes in the last two minutes, but Miami kept answering. And then UVA missed two FTs with 5 seconds left, and Miami's Angel Rodriguez got fouled shooting a 3 with less than a second left. He hit all three FTs, despite UVA trying to ice him after the second one. To overtime we go. Shane Battier on the call.

[pfrduke]

If Duke is equal or less than 3 losses would that be considered as the best regular season ever?

1986 and 1999 would like a word. 1992 as well.

[arnie]

UVA going to OT against Miami. Gritty performance by Miami. I think they were down 38-20 at or maybe shortly after halftime, but from that point they quickly brought the game within 10 and then kept grinding. Virginia hit two big threes in the last two minutes, but Miami kept answering. And then UVA missed two FTs with 5 seconds left, and Miami's Angel Rodriguez got fouled shooting a 3 with less than a second left. He hit all three FTs, despite UVA trying to ice him after the second one. To overtime we go. Shane Battier on the call.

Big deal in the reg season race if it goes Miami.

[Merlindevildog91]

Double OT now.

We watched the first half and then went to the Panthers game, thinking this one was in the bag early for UVA. Guess I thought wrong.

[Mabdul Doobakus]

I think Battier jinxed Miami. For some reason, he predicted a Miami win after Miami went up 3 in this OT. It's been 6-0 UVA since then. Blame Battier.

[W&LHoo]

Virginia missed a ton of free throws - clearly need to work on that. Didn't help that Brogdon fouled out either.

Also, the team got a bit cocky after being up 18 on the road at the half and got a good message. Second test in a row for my hoos, and they came through. I expect they'll be ready to gut out the games against ND and you guys before we see you.

[uh_no]

Virginia missed a ton of free throws - clearly need to work on that. Didn't help that Brogdon fouled out either.

Also, the team got a bit cocky after being up 18 on the road at the half and got a good message. Second test in a row for my hoos, and they came through. I expect they'll be ready to gut out the games against ND and you guys before we see you.

Not winning decisively does not mean you have "guts," it means you're not good enough to win decisively. If only there were a recent example of a highly lauded team "gutting out wins" against mediocre competition before getting absolutely smoked like a brat against actual competition.

interestingly enough, Ken-pom measures the difference between the number of wins you should have based on your teams abilities vs your actual win total. It can be viewed as a measure of "gutty-ness" if you like....but he calls it "luck," which is likely more faithful to the actual phenomenon. By this metric, duke is twice as gutty as UVA is :)

[W&LHoo]

In this instance, by "gut out" I really meant "maintain focus and execution." I'll endeavor to post more precisely in the future :)

What we had in this game is a Virginia team that dominated the 1st half and laid off the gas. Miami got on a run, got the crowd involved, and rattled our guys. That is something that isn't acceptable in league play, and the last two games should emphasize that for my hoos.

[jv001]

Clearly Anderson went from Goat to Hero with the 3 that tied the game in the first OT. What got me when he committed the dumb foul, he acted like he didn't hit the shooter's elbow. Lucky win, but it seems like the good teams are lucky,lol. GoDuke!

[Olympic Fan]

Anderson was clutch on OT, but I really think his three FTs on the 3 were bogus -- he used an old Jon Scheyer trick, kicking out his legs to initiate contact.

So Virginia escapes to remain unbeaten, but Colorado State goes down at New Mexico. Villanova and TCU beaten earlier in the day.

That leaves us with three unbeaten -- Kentucky, Duke and Virginia.

PS I found an RPI site earlier in the day. It's just getting to time where RPI means something. TCU was No. 97 BEFORE losing today. Their perfect record was about as bogus as its gets.

[devildeac]

Anderson was clutch on OT, but I really think his three FTs on the 3 were bogus -- he used an old Jon Scheyer trick, kicking out his legs to initiate contact.

So Virginia escapes to remain unbeaten, but Colorado State goes down at New Mexico. Villanova and TCU beaten earlier in the day.

That leaves us with three unbeaten -- Kentucky, Duke and Virginia.

PS I found an RPI site earlier in the day. It's just getting to time where RPI means something. TCU was No. 97 BEFORE losing today. Their perfect record was about as bogus as its gets.

What? More bogus than upper level AFAM/PHIL/(fill in name here) courses at c*rolina?:rolleyes:

[duketaylor]

"If Duke is equal or less than 3 losses would that be considered as the best regular season ever?"

Not sure why this question as Duke's had a team that went 37-3, but didn't win a NC, meaning they only had two regular-season losses.

Any regular-season record will be skewed by post-season success. Must win a NC for regular-season to be truly meaningful, at least to me. If I had to endure 7 or 8 losses before a NC, I'd take that over a one-loss non-NC team.

Anybody else feel that way?

Ask Wolfpack fans about 1983, bet they'd agree.

[Wander]

Not winning decisively does not mean you have "guts," it means you're not good enough to win decisively. If only there were a recent example of a highly lauded team "gutting out wins" against mediocre competition before getting absolutely smoked like a brat against actual competition.

interestingly enough, Ken-pom measures the difference between the number of wins you should have based on your teams abilities vs your actual win total. It can be viewed as a measure of "gutty-ness" if you like....but he calls it "luck," which is likely more faithful to the actual phenomenon. By this metric, duke is twice as gutty as UVA is :)

If it's a consistent trend, then sure. But if it's just a couple times, then I don't think it means the team isn't good enough to win decisively. Most if not all teams have close games against inferior competition at some point - it doesn't mean that they're FSU.

Gutty-ness and Luck are two explanations for the stat that kenpom measures, but I prefer a third one: The Degree To Which Kenpom's Metric Fails For That Team.

[uh_no]

If it's a consistent trend, then sure. But if it's just a couple times, then I don't think it means the team isn't good enough to win decisively. Most if not all teams have close games against inferior competition at some point - it doesn't mean that they're FSU.

Gutty-ness and Luck are two explanations for the stat that kenpom measures, but I prefer a third one: The Degree To Which Kenpom's Metric Fails For That Team.

I disagree that it constitutes failure.

Think of this analogy:

When rolling 2 dice, we know there is a 1/6 chance of rolling 7. If you rolled 6 7's in a row, and only ever rolled the dice 6 times, would you say that the probability distribution metric failed to explain that dice?

Obviously the probability says you should only roll 1 7, but you rolled 6!!! so who is lying????

The metric didn't fail...it actually says that 1/46656 times you WILL roll 6 7's.

What you CAN say, however, after rolling the dice is that the rolls deviated a large amount from the most expected results.

It holds similar for kenpom. Consider the probability distribution to be his efficiencies (which are quite literally directly converted to probability distributions...it's what the "pyth" is), and the dice roll governed by that distrbution the actual games. The system doesn't fail if a team has a poor record despite great efficiencies, it simply says that this is not the most likely result given the actual efficiencies, same as rolling 6 7's isn't the most likely result given the probability distribution.

Rolling 6 7's would be considered "lucky". Rolling 0 7's would be considered unlucky. The distribution didn't fail in either case....we just ended up in the tail of the bell curve.

While it is true that the kenpom efficiencies at any given point in time suffer from an extreme lack of data points (unlike the dice rolls), meaning that they may end up being demosntrated inaccurate down the line, the assumption when calculating luck is that they are correct. Further, even if you DID have well-known efficiences that were correct, there would STILL be a luck factor....as every team wouldn't hit their exact efficiency in every game, just like the dice doesn't hit it's distribution exactly every 6 rolls. This is where the law of large numbers comes in.

In theory if a team played every game 1000 times (instead of once), the luck for every team would start to approach 0, as the win% of a team in those 1000 games should closely align with what is predicted by the distribution derived from the efficiencies, much like if you rolled the dice a gazillion times, you would get 7 once every 6 times.

Ultimately luck stems from small N trials, not a failure of the system.

[Wander]

I disagree that it constitutes failure.

The purpose of the kenpom system is not to explain history, but to provide a likelihood of some event happening in the future. Think of this analogy:

When rolling 2 dice, we know there is a 1/6 chance of rolling 7. If you rolled 6 7's in a row, and only ever rolled the dice 6 times, would you say that the probability distribution metric failed to explain that dice?

Obviously the probability says you should only roll 1 7, but you rolled 6!!! so who is lying????

The metric didn't fail...it actually says that 1/46656 times you WILL roll 6 7's.

What you CAN say, however, after rolling the dice is that the rolls deviated a large amount from the most expected results.

It holds similar for kenpom. Consider the probability distribution to be his efficiencies (which are quite literally directly converted to probability distributions...it's what the "pyth" is), and the dice roll governed by that distrbution the actual games. The system doesn't fail if a team has a poor record despite great efficiencies, it simply says that this is not the most likely result given the actual efficiencies, same as rolling 6 7's isn't the most likely result given the probability distribution.

Rolling 6 7's would be considered "lucky". Rolling 0 7's would be considered unlucky. The distribution didn't fail in either case....we just ended up in the tail of the bell curve.

But you're using an example where we know for certain the probability distribution beforehand, and it's easy to calculate. If you roll a die and kept getting 6 over... and over... and over... again, at some point you would in fact question if the die was weighted and that your idea for its 1-in-6 probability distribution was incorrect. Or: if you had a die that you knew might be weighted, you wouldn't be quick to assume that rolling seven 6's in a row was just lucky.

You won't find anyone who loves kenpom more than me, but is it really that hard to believe that the system contains imperfections? (I've long suspected that he overrates slow teams, which I will one day draw up a graph or two for and actually test). Any systematic imperfections should, in fact, show up in his luck ratings.

[Acymetric]

Wander beat me to it, but just to reaffirm in a much shorter way, the problem with the dice example is that basketball games are not a roll of the dice. Looking for a good non-sports statistical corollary, the closest might be meteorology? At least in the "this is what has happened in the past so here is probably what will happen this time" sense. In both cases the predicted outcome doesn't always match reality.

[Jim3k]

If Duke is equal or less than 3 losses would that be considered as the best regular season ever?

The short answer is no. In 1998-99, Duke's regular season was 29-1, the only loss to non-conference Cincinnati. .977 win rate.

It did not lose again until the National Championship game. Overall record was 37-2 (http://www.sports-reference.com/cbb/schools/duke/1999-schedule.html), a .948 win record.

[burnspbesq]

So-called "advanced metrics" are interesting and fun, but their predictive power is somewhat overrated, and that's true of every sport in which their use has become prevalent. Certainly Duke fans should understand this. None of the last five Duke teams to win national championships (women's golf 2014, men's lax 2010/13/14, men's basketball 2010) has gotten even a sniff of the top of any "advanced" ranking. Whatever it is that causes a number four player to hit the shot of her life when her team needs it most, or causes a goalie to play terribadly for 57 minutes and then make a ridiculous save and an even better outlet pass, isn't capured by advanced metrics.

And on some level, isn't tnat why we watch?

[uh_no]

But you're using an example where we know for certain the probability distribution beforehand, and it's easy to calculate. If you roll a die and kept getting 6 over... and over... and over... again, at some point you would in fact question if the die was weighted and that your idea for its 1-in-6 probability distribution was incorrect. Or: if you had a die that you knew might be weighted, you wouldn't be quick to assume that rolling seven 6's in a row was just lucky.

You won't find anyone who loves kenpom more than me, but is it really that hard to believe that the system contains imperfections? (I've long suspected that he overrates slow teams, which I will one day draw up a graph or two for and actually test). Any systematic imperfections should, in fact, show up in his luck ratings.

Right, but that's not what luck demonstrates....as I mentioned, you would STILL have a non-zero luck even if you DID have perfect efficiencies (as with the dice).

You're ultimately right in that there's no way to differentiate between a too-small sample size and an incorrect model....with the dice we set a threshhold and do silly stuff like have a null hypothesis and a confidence interval, and then roll the dice, and if we get too many we can say that there is enough evidence to reject the null hypothesis...T test or something like that. But we can roll the dice as many times as we want to know for sure.

We don't have the option to play more games, but the base assumption of ALL his calculations (probabilities, SOS, pyth) is that the current model is correct. Without that assumption, we could make the same argument to reject ALL kenpom stats...

By using any kenpom at all, you're making the implied assumption that there is enough data in the season to come up with reasonable efficiency numbers. And if you accept that assumption that the efficiency numbers are reasonable, then you accept that luck is simply the degree to which a teams record is likely. If you don't accept kenpom's efficiencies as reasonable, though, then why are you using kenpom stats in the first place?

THe measure of whether an individual team is failed by the system is much more accurately reflected in the error in predicting game by game efficiencies. Here's an example:

Under the current ratings, BC should have put up a 90 or so on duke. In reality they put up an 98. That's a pretty big error. That represents the system's inability to perfectly align all results....as I mentioned in another post, is it that BC is a better offensive team? that duke is a better defensive team? or just the tail end of the bell curve?

All we know is it's an error, and one that the system simply chalks up to a probability distribution of an individual game efficiency.
The fact that the system erred there, however, does NOT show up in luck....since duke won the game anyway, and the system said they were going to win the game.

So just a small example where error in the system is better measured without taking records into account.

Aside: you must accept that a team's score is determined by their efficiency in a given game....which is easy to do...since efficiency*tempo=final score, and that their game efficiency is some team-by-team efficiency distribution based on a mean and standard deviation

So if efficiency is probabilistic within a game, then the win probability is also probabilistic (truism?). Therefore if we accept the error in the efficiency calculation as amount that the system failed to explain a given result in terms of efficiency, there is no reason to ALSO calculate how if failed to predict a win or loss....since the WL record error would be directly calculable from the efficiency error (error*tempo=final error in score).

I think the point is that before we even begin to calculate Luck, the system has either failed or succeeded....the efficiency values are either correct, or they're not...and there's no way to know. All we know is our errors in efficiency, which may or may not arise in Luck as well (depending on whether the team actually won the game). Luck does not give you any NEW information on whether the efficiencies are correct or not, after considering the errors, and therefore is only useful if you ACCEPT The "errors" in the efficiencies as explained as simply random variance.

That acceptance is equivalent to accepting the efficiencies as correct....or flipped around, if you don't accept the efficiencies, then you can't accept the luck value as meaningful either.

[Olympic Fan]

If Duke is equal or less than 3 losses would that be considered as the best regular season ever?

The best regular season in Duke history is CLEARLY 1999 -- 29-1 with a one-point loss to Cincinnati (which basically had the same team as would spend most of the 2000 season at No. 1) in Alaska, when half the team was battling the flu. That team beat five top 10 teams in the regular season (two on the road and two on neutral courts). 26 of the 29 wins were by double digits. And the team finished at No. 1.

THAT is the standard 2015 will have to beat to have the best regular season ever.

The second best regular season was 1986 -- 29-2 and No. 1 in the nation. They beat 11 ranked teams (four top 10) and won 21 of those 29 games by double digits.

The third best regular season was 1992 -- 28-2 and No. 1 every week of the season. Beat nine ranked teams (three top 10; it doesn't count a road win at the unranked Michigan team that would play for the national title) and won 18 games by double digits.

Two other teams have finished the regular season with two losses -- Bubas' 1963 team (Heyman's senior year) was 22-2 before the ACC Tournament and Eddie Cameron's 1942 team was 19-2

You might want to count the ACC Tournament as regular season (pre-NCAA). All of those teams won conference tournaments. The 1999 team won three games by 37 points, by 15 points and by 23 points, so that would make them 32-1 with 29 double digit wins before the NCAA. The 1986 team would be 32-2; the 1992 team 31-2, the '63 team 25-2 and the '41 team 22-2 (which was in fact the final record -- not sue how they missed a NCAA bid that year).

[Wander]

So-called "advanced metrics" are interesting and fun, but their predictive power is somewhat overrated, and that's true of every sport in which their use has become prevalent. Certainly Duke fans should understand this. None of the last five Duke teams to win national championships (women's golf 2014, men's lax 2010/13/14, men's basketball 2010) has gotten even a sniff of the top of any "advanced" ranking. Whatever it is that causes a number four player to hit the shot of her life when her team needs it most, or causes a goalie to play terribadly for 57 minutes and then make a ridiculous save and an even better outlet pass, isn't capured by advanced metrics.


I don't know about the other examples but that's definitely not true for Duke basketball in 2010, in which advanced metrics had Duke on top in a season when the common wisdom was that we were worse than Kansas, Kentucky, and Syracuse.

[pfrduke]

So-called "advanced metrics" are interesting and fun, but their predictive power is somewhat overrated, and that's true of every sport in which their use has become prevalent. Certainly Duke fans should understand this. None of the last five Duke teams to win national championships (women's golf 2014, men's lax 2010/13/14, men's basketball 2010) has gotten even a sniff of the top of any "advanced" ranking. Whatever it is that causes a number four player to hit the shot of her life when her team needs it most, or causes a goalie to play terribadly for 57 minutes and then make a ridiculous save and an even better outlet pass, isn't capured by advanced metrics.

And on some level, isn't tnat why we watch?

The Men's Basketball team was either the #1 or #2 Pomeroy ranked team going into the NCAA tournament. The continued high ranking of the Blue Devils that season was a topic of some derision (http://kenpom.com/blog/index.php/weblog/entry/one_shining_e-mail) from "eye test" observers.

That said, I agree with the overall tenor of your observation. Any rating system (computers, human polls, Vegas oddsmakers, etc.) is not particularly useful as predictive tool for the outcome of a single game.

[Wahoo2000]

PS I found an RPI site earlier in the day. It's just getting to time where RPI means something. TCU was No. 97 BEFORE losing today. Their perfect record was about as bogus as its gets.

Something maybe, but not much. That's as flawed a mathematical standard as there is in team sports. I strongly prefer the predictive tools like Kenpom, Sagarin, etc. It drives me nuts that the selection committee's decisions seem to often follow RPI more closely that any other metric when it comes to deciding which teams are in and out.

And apologies - in this case, and for the example you were making in regard to TCU being overrated/overhyped due to their record, RPI was fine. I just go crazy almost anytime RPI is brought up.

[Des Esseintes]

Not winning decisively does not mean you have "guts," it means you're not good enough to win decisively. If only there were a recent example of a highly lauded team "gutting out wins" against mediocre competition before getting absolutely smoked like a brat against actual competition.

interestingly enough, Ken-pom measures the difference between the number of wins you should have based on your teams abilities vs your actual win total. It can be viewed as a measure of "gutty-ness" if you like....but he calls it "luck," which is likely more faithful to the actual phenomenon. By this metric, duke is twice as gutty as UVA is :)

Can we talk for a second about kenpom's luck measure? It really befuddles me in some ways, and Duke vs. UVA is a case in point. Both teams have played 13 games, but his systems finds that UVA has played a significantly tougher SOS than Duke, .5857 to . 5345. Meanwhile Duke and UVA have almost identical power ratings, .9708 to .9689. Now, given that the system presently thinks the two teams are almost equivalently good while UVA has played the superior schedule, shouldn't it then conclude that UVA has been luckier to reach its 13-0 record? But it doesn't. It thinks Duke has been quite a bit luckier this season. Does that make sense to anyone?

Another example. Louisville and Arizona are both 12-1, and the system rates Louisville as superior. It also believes Louisville has played the weaker SOS. So why does it think Louisville, the stronger team playing the easier schedule, is luckier to get to the same record as Arizona?

This has been happening for years. I'm sure there is an obvious explanation I am overlooking, but, well, I have overlooked it. Can someone enlighten me?

[bob blue devil]

Can we talk for a second about kenpom's luck measure? It really befuddles me in some ways, and Duke vs. UVA is a case in point. Both teams have played 13 games, but his systems finds that UVA has played a significantly tougher SOS than Duke, .5857 to . 5345. Meanwhile Duke and UVA have almost identical power ratings, .9708 to .9689. Now, given that the system presently thinks the two teams are almost equivalently good while UVA has played the superior schedule, shouldn't it then conclude that UVA has been luckier to reach its 13-0 record? But it doesn't. It thinks Duke has been quite a bit luckier this season. Does that make sense to anyone?

Another example. Louisville and Arizona are both 12-1, and the system rates Louisville as superior. It also believes Louisville has played the weaker SOS. So why does it think Louisville, the stronger team playing the easier schedule, is luckier to get to the same record as Arizona?

This has been happening for years. I'm sure there is an obvious explanation I am overlooking, but, well, I have overlooked it. Can someone enlighten me?

that caught my eye too. i'll throw a dart at it and see if it sticks...

i'm guessing it relates to the fact that we've had the toughest game by a wide margin - at wisconsin. kenpom says wisconsin was the favorite in that game and if we hadn't had "good luck" (by kenpom's definition), we would have lost. whereas, uva was the favorite in each of its games and didn't need "good luck" in any particular, despite playing a tougher schedule overall. to simplify the math, if you are an average team and have two games where you are 50/50 in each (so an average SoS), then your chances of winning both are 25%. now change that to playing the same overall "SoS", but one game against a superior opponent and one against a lesser one - where you are only 25% likely to beat the superior one and 75% likely to beat the lesser. your team is still average and your strength of schedule hasn't changed, but your odds of winning both are lower (18.75%) - so if you won both you've had better luck. it's almost like a geometric mean vs. an arithmetic one. at least that's my guess...

[freshmanjs]

Can we talk for a second about kenpom's luck measure? It really befuddles me in some ways, and Duke vs. UVA is a case in point. Both teams have played 13 games, but his systems finds that UVA has played a significantly tougher SOS than Duke, .5857 to . 5345. Meanwhile Duke and UVA have almost identical power ratings, .9708 to .9689. Now, given that the system presently thinks the two teams are almost equivalently good while UVA has played the superior schedule, shouldn't it then conclude that UVA has been luckier to reach its 13-0 record? But it doesn't. It thinks Duke has been quite a bit luckier this season. Does that make sense to anyone?

Another example. Louisville and Arizona are both 12-1, and the system rates Louisville as superior. It also believes Louisville has played the weaker SOS. So why does it think Louisville, the stronger team playing the easier schedule, is luckier to get to the same record as Arizona?

This has been happening for years. I'm sure there is an obvious explanation I am overlooking, but, well, I have overlooked it. Can someone enlighten me?

it's not just about record. margin of victory matters too.

[bob blue devil]

kenpom has us the big underdog of the 3 remaining teams:
- duke at 1.4%
- uva at 4.9%
- kentucky at 27.2%

a very large chunk of the difference between duke and uva is simply playing at uva. unbalanced schedule didn't do us any favors this year! we play away at the acc's top 4 teams (top 6 if you go by kenpom's ratings and bump miami below syracuse and nc state)...

kentucky has an interesting road - no singular hurdles left (post louisville), but a handful of potentially game opponents on the road like florida, south carolina, georgia and lsu. someone needs to get hot from long range against these guys.

[Des Esseintes]

it's not just about record. margin of victory matters too.

Margin of victory--or, more accurately, a team's adjusted efficiency--goes into a team's power ranking. That's not a value outside kenpom's numbers.


that caught my eye too. i'll throw a dart at it and see if it sticks...

i'm guessing it relates to the fact that we've had the toughest game by a wide margin - at wisconsin. kenpom says wisconsin was the favorite in that game and if we hadn't had "good luck" (by kenpom's definition), we would have lost. whereas, uva was the favorite in each of its games and didn't need "good luck" in any particular, despite playing a tougher schedule overall. to simplify the math, if you are an average team and have two games where you are 50/50 in each (so an average SoS), then your chances of winning both are 25%. now change that to playing the same overall "SoS", but one game against a superior opponent and one against a lesser one - where you are only 25% likely to beat the superior one and 75% likely to beat the lesser. your team is still average and your strength of schedule hasn't changed, but your odds of winning both are lower (18.75%) - so if you won both you've had better luck. it's almost like a geometric mean vs. an arithmetic one. at least that's my guess...

This makes sense. If Duke had played the Spurs instead of Toledo and remained undefeated, it would have a significantly stronger schedule strength and a waaaaaay larger magnitude of luck.

[Des Esseintes]

kenpom has us the big underdog of the 3 remaining teams:
- duke at 1.4%
- uva at 4.9%
- kentucky at 27.2%

a very large chunk of the difference between duke and uva is simply playing at uva. unbalanced schedule didn't do us any favors this year! we play away at the acc's top 4 teams (top 6 if you go by kenpom's ratings and bump miami below syracuse and nc state)...

kentucky has an interesting road - no singular hurdles left (post louisville), but a handful of potentially game opponents on the road like florida, south carolina, georgia and lsu. someone needs to get hot from long range against these guys.

On the other hand, ken argues that unbalanced schedules almost never decide (http://kenpom.com/blog/index.php/weblog/entry/on_unbalanced_conference_schedules) conference champions. A single game's difference in the win-loss column is considered anomalous, and a two game difference, he says, is perhaps impossible.



It’s true that the best team doesn’t always win its conference title. And in rare cases, the schedule can swing a conference race to one team or the other. But the biggest factor in preventing the best team from winning its conference is more often going to be the random things that influence the outcome of close games.

[vick]

On the other hand, ken argues that unbalanced schedules almost never decide (http://kenpom.com/blog/index.php/weblog/entry/on_unbalanced_conference_schedules) conference champions. A single game's difference in the win-loss column is considered anomalous, and a two game difference, he says, is perhaps impossible.

I agree with the thrust of Ken's post there. Frankly I've always found the way he posts conference strength of schedule to be a little strange intuitively, since it's just the average ranking of teams you play--so even a double round robin would be "unbalanced" since the better teams don't have to play themselves--and indeed when you look at it through the years, being a top team is significantly correlated with a "weaker" schedule. Is Duke's schedule really "weaker" than Virginia Tech's because it doesn't play itself? The second issue is that it ignores home court advantage, which matters when you are translating "Pomeroy strength" to "Win-loss strength"--i.e., a schedule is more favorable from a win-loss perspective if your home court games are against teams of similar ability (the easiest way for me to think about this was to imagine that home court didn't exist, but you could add three points to your score at the beginning of half of your games. You wouldn't select them randomly!).

I took a look at this using a slightly different methodology, where I took the expected number of wins for each team using Pomeroy's offensive and defensive efficiencies against their actual schedule. I then compared this to what the winning percentage would be against a double round robin, and then scaled that to an 18 game schedule (i.e., multiplied by 18/28, since there would be 28 games in a double round robin). I got this for each team:


Team Expected Wins "Fair" Expected Wins Overall schedule advantage
Virginia 15.3 15.4 0.0
Duke 14.9 15.2 -0.3
Louisville 13.2 13.6 -0.4
North Carolina 12.6 13.0 -0.4
Notre Dame 11.1 11.0 0.1
Syracuse 9.9 9.8 0.1
NC State 8.0 8.3 -0.2
Pittsburgh 8.1 7.9 0.3
Miami FL 8.2 7.8 0.4
Georgia Tech 7.4 7.2 0.2
Wake Forest 5.3 5.9 -0.6
Boston College 5.2 5.5 -0.3
Florida St. 6.1 5.4 0.7
Clemson 5.5 5.1 0.4
Virginia Tech 3.9 3.9 0.0


Not really all that meaningful a difference. By the way, Duke's a good example of why you have to look at home court in addition to who you play in an unbalanced schedule, which Pomeroy's SOS ignores--having to play Louisville and Virginia away pretty much accounts for all of the difference between Duke's actual schedule difficulty and a "fair" schedule (from a "who you play" perspective, Duke's relatively neutral this year since Wake is not good and they're one of the four teams we play twice). This methodology has its flaws as well, particularly at the extremes of very good or very bad teams, but I think it's some evidence that schedule unbalance is probably not something to spend time worrying about.

[bob blue devil]

On the other hand, ken argues that unbalanced schedules almost never decide (http://kenpom.com/blog/index.php/weblog/entry/on_unbalanced_conference_schedules) conference champions. A single game's difference in the win-loss column is considered anomalous, and a two game difference, he says, is perhaps impossible.

humorously enough i read that post by him and wasn't fully sold because it ignores... wait for it... luck! at least the conclusion (that schedules almost never decide conference champions) isn't fully proven by the argument. he makes the argument in expectation, but ignores the randomness in outcomes (he only answers the question of whether the favorite is still the favorite). IMO it should've been put together in terms of probability of winning conference title vs. home and home round robin schedule (i.e. duke has an x% chance of winning the title given current schedule make-up vs. a y% chance had everyone in the acc played a home and home round robin). then we would have a better guess as to what kenpom's system predicts the impact of an unbalanced schedule has on a team's title hopes. w.r.t. duke and the acc - i can't imagine it has a totally negligible impact in expectation as duke and uva are almost the same in his system and playing at uva (vs. neutral) gives them an extra ~17% chance of winning the game (so, all else equal, uva has an ~0.3 game heads start).

[Olympic Fan]

On the other hand, ken argues that unbalanced schedules almost never decide (http://kenpom.com/blog/index.php/weblog/entry/on_unbalanced_conference_schedules) conference champions. A single game's difference in the win-loss column is considered anomalous, and a two game difference, he says, is perhaps impossible.

The key word is almost -- it does happen. It happened in the ACC in 2013. I forget the exact circumstances now, but I know that Duke and Miami played something like 13 identical games -- and Duke had a better record in those games. But Miami won the title because of the five easier games that they got to play.

And Wahoo, I agree that RPI is a flawed metric (I much prefer Pomeroy), but it's the one that carries the most weight with the selection committee, so we ignore it at risk.

[uh_no]

that caught my eye too. i'll throw a dart at it and see if it sticks...

i'm guessing it relates to the fact that we've had the toughest game by a wide margin - at wisconsin. kenpom says wisconsin was the favorite in that game and if we hadn't had "good luck" (by kenpom's definition), we would have lost. whereas, uva was the favorite in each of its games and didn't need "good luck" in any particular, despite playing a tougher schedule overall. to simplify the math, if you are an average team and have two games where you are 50/50 in each (so an average SoS), then your chances of winning both are 25%. now change that to playing the same overall "SoS", but one game against a superior opponent and one against a lesser one - where you are only 25% likely to beat the superior one and 75% likely to beat the lesser. your team is still average and your strength of schedule hasn't changed, but your odds of winning both are lower (18.75%) - so if you won both you've had better luck. it's almost like a geometric mean vs. an arithmetic one. at least that's my guess...

That's precisely correct. If a team has a 100% chance to win every game, and wins every game, that's 0 luck. If a team has a 50.00001% chance to win every game, and wins 10 games, despite being favored, that's probably something like .5 luck. So since duke had only a small chance to beat wisconsin, even if Duke WAS favored, it takes significantly more luck than it would to beat Grambling state.

So effectively the luck metric is dominated by performances against teams with whom you were projected to have a close game. Win those games, and you're lucky. lose them, and you're unlucky.

FSU in football this year, for example, would have been considered stupendously lucky.

As for what Kenpom's physical quantity actually represents, I believe it's the amount which you'd have to add to their pythagorean winning percentage for it to have made their record explainable. As I mentioned, as you play a huge number of games, your record determined by efficiencies should match your actual record, and luck would approach zero. (because if we had a 75% chance of beating wisconsin, we would lose once out of every four times....negating the positive luck of beating them after only having played once)

[Wander]

As I mentioned, as you play a huge number of games, your record determined by efficiencies should match your actual record, and luck would approach zero.

My issue is that this statement is only true if the efficiencies are perfect for every team. What I'm contending is that they are not, and would not be even if the season had 10000 games for each team. Thus my statement that "luck" actually measures the imperfections in his system, in addition to small sample size issues.

[uh_no]

humorously enough i read that post by him and wasn't fully sold because it ignores... wait for it... luck!

Luck is the premise of the whole article. The thesis is that the variance in outcomes (luck) is far more important in determining a conference winner than schedule. He literally says that


But the biggest factor in preventing the best team from winning its conference is more often going to be the random things that influence the outcome of close games.

The problem is, as you point out, that when he goes to do the actual analysis, he tosses this out the window. He looks at the final outcome as predicted by his efficiency and his conference rankings...while 100% ignoring the fact that while, say, the ACC's order is the same, the likelihood of the outcomes would change....maybe in a balanced schedule duke would have a 60% chance, but with unbalanced it would be 30%! I'd say that matters a great deal! once out of every 3 years!

It would be much better to come up with some metric that would examine the changes in likelihood of moving up or down a position because of unbalancedness.

[vick]

It would be much better to come up with some metric that would examine the changes in likelihood of moving up or down a position because of unbalancedness.

This seems like it would be pretty tough to do precisely, though, because the unbalanced schedule is also shorter than the balanced schedule (18 vs. 28 games), which introduces more variance in the results. Maybe I'll try to dust off some VB macros (it's only been 5+ years since I used them...) and see if I can't simulate the results from both schedules to see how different they are. I'm not even sure how you would address the length-of-season difference though.

[uh_no]

This seems like it would be pretty tough to do precisely, though, because the unbalanced schedule is also shorter than the balanced schedule (18 vs. 28 games), which introduces more variance in the results. Maybe I'll try to dust off some VB macros (it's only been 5+ years since I used them...) and see if I can't simulate the results from both schedules to see how different they are. I'm not even sure how you would address the length-of-season difference though.

It shouldn't matter. All we're looking for is a probability that a team finishes in any given position, and that should just require a gazillion simulations of a given schedule.

[jdk]

The best regular season in Duke history is CLEARLY 1999 -- 29-1 with a one-point loss to Cincinnati (which basically had the same team as would spend most of the 2000 season at No. 1) in Alaska, when half the team was battling the flu. That team beat five top 10 teams in the regular season (two on the road and two on neutral courts). 26 of the 29 wins were by double digits. And the team finished at No. 1.

THAT is the standard 2015 will have to beat to have the best regular season ever.


That was the most dominant Duke basketball team I can remember. Against Cincinnati, I believe we simply ran out of time...Will Avery had either a tying shot or a game-winning shot before overtime that was waived off because the clock expired. The <10 point victories were 2x against next year's champion Michigan State (6 points each time), @Ga Tech was an 8 point victory, and the St John's game.

Regarding undefeated, Duke was only half a second late against Cincinnati, and a Trajan Langdon travel call against UConn away.

I also remember JJ Redick's senior season, we started out something like 17-0 or 18-0, and lost at Georgetown. That was the year that we blew out Texas in MSG.

1999 is the gold standard for a regular season IMHO, and this year is the closest I can remember to feeling that way. Obviously we have a long way to go.

[bob blue devil]

This seems like it would be pretty tough to do precisely, though, because the unbalanced schedule is also shorter than the balanced schedule (18 vs. 28 games), which introduces more variance in the results. Maybe I'll try to dust off some VB macros (it's only been 5+ years since I used them...) and see if I can't simulate the results from both schedules to see how different they are. I'm not even sure how you would address the length-of-season difference though.

could we make a statistical inference about the distribution of outcomes for an only possible in theory balanced 18 game schedule using the results of the 28 game balanced schedule? if they were normally distributed, easy as pie, but this foe is beyond my powers.

[CDu]

My issue is that this statement is only true if the efficiencies are perfect for every team. What I'm contending is that they are not, and would not be even if the season had 10000 games for each team. Thus my statement that "luck" actually measures the imperfections in his system, in addition to small sample size issues.

It is true that Pomeroy is using the term luck to describe the unexplained error between the actual and the predicted values. It is impossible to differentiate what part of that error is luck and what is flaw in the model. Hence uh no's point that you either accept the model as correct (meaning error is really luck) or you don't (meaning it is not). In reality, it is almost certainly a mix of luck and flaw. But we don't know how much of each. So the best one can do is either accept the model or find ways to improve it. But ultimately because we are constantly dealing with a small sample size relative to the number of variables, it is probably impossible to really determine whether the model is close to perfect or not.

[vick]

could we make a statistical inference about the distribution of outcomes for an only possible in theory balanced 18 game schedule using the results of the 28 game balanced schedule? if they were normally distributed, easy as pie, but this foe is beyond my powers.

This, unfortunately, is beyond the skills that I dimly remember from Stat 103 a decade ago, but I think I was able to successfully simulate 10,000 seasons with the real schedule and a 28-game double round robin schedule, and here are the chances of winning the conference title (I removed teams with very low chances for the sake of readability):





Virginia
Duke
Louisville
UNC
Notre Dame
Syracuse


Real schedule
Outright
40.5%
24.9%
5.6%
3.0%
0.3%
0.1%



Tied
21.0%
19.6%
8.8%
5.2%
1.0%
0.3%




Combined
61.5%
44.5%
14.4%
8.3%
1.3%
0.3%


Round robin
Outright
40.0%
34.2%
4.5%
2.0%
0.1%
0.0%



Tied
16.3%
15.9%
5.3%
3.1%
0.3%
0.0%



Combined
56.3%
50.1%
9.9%
5.1%
0.3%
0.0%



As you'd expect,the combined percentages go down for the round robin (more games means less chance of a tie at the top). You can see how Duke's schedule is harder than UVa's by this metric, and how the secondary contenders--Louisville and UNC in particular--are helped out by the shorter schedule even though their schedules aren't demonstrably easier than Duke's (UVa's is). Of course I think these probabilities probably overstate Duke and UVa's real chances since they are based on performance to-date and therefore don't account for things like potential injuries, etc., but maybe this gives an idea of the magnitude of the impact of schedule differences.

[uh_no]

This, unfortunately, is beyond the skills that I dimly remember from Stat 103 a decade ago, but I think I was able to successfully simulate 10,000 seasons with the real schedule and a 28-game double round robin schedule, and here are the chances of winning the conference title (I removed teams with very low chances for the sake of readability):


my only regret is that I have but one spork to give for this post.

great job. this is literally infinitely more helpful than kenpoms blog post. great job.

[Wander]

It is true that Pomeroy is using the term luck to describe the unexplained error between the actual and the predicted values. It is impossible to differentiate what part of that error is luck and what is flaw in the model. Hence uh no's point that you either accept the model as correct (meaning error is really luck) or you don't (meaning it is not). In reality, it is almost certainly a mix of luck and flaw. But we don't know how much of each. So the best one can do is either accept the model or find ways to improve it. But ultimately because we are constantly dealing with a small sample size relative to the number of variables, it is probably impossible to really determine whether the model is close to perfect or not.

I agree that his "luck" is actually a mix of luck and model flaws to some degree. I do think it's possible to get some sort of idea on how much of each is present. For example: my guess is that end of game situations involving intentional fouling don't work very well in a strict pace-adjusted system, which would cause slow teams that are good at free throw shooting (like Wisconsin and Virginia) to be slightly overrated. This is something that should be relatively simple to test.

Another example is that I don't think it's unreasonable to suggest that some teams are genuinely better at winning close games. Even if you don't buy into clutch shooting (which I don't), you can still have effects like teams insisting that Kobe Bryant always take the last shot in a one possession game, which decreases the chances of winning while having only a very minor effect on efficiency numbers. Which is admittedly a little harder to test for 351 teams.

[Kfanarmy]

This, unfortunately, is beyond the skills that I dimly remember from Stat 103 a decade ago, but I think I was able to successfully simulate 10,000 seasons with the real schedule and a 28-game double round robin schedule, and here are the chances of winning the conference title (I removed teams with very low chances for the sake of readability):





Virginia
Duke
Louisville
UNC
Notre Dame
Syracuse


Real schedule
Outright
40.5%
24.9%
5.6%
3.0%
0.3%
0.1%



Tied
21.0%
19.6%
8.8%
5.2%
1.0%
0.3%




Combined
61.5%
44.5%
14.4%
8.3%
1.3%
0.3%


Round robin
Outright
40.0%
34.2%
4.5%
2.0%
0.1%
0.0%



Tied
16.3%
15.9%
5.3%
3.1%
0.3%
0.0%



Combined
56.3%
50.1%
9.9%
5.1%
0.3%
0.0%



As you'd expect,the combined percentages go down for the round robin (more games means less chance of a tie at the top). You can see how Duke's schedule is harder than UVa's by this metric, and how the secondary contenders--Louisville and UNC in particular--are helped out by the shorter schedule even though their schedules aren't demonstrably easier than Duke's (UVa's is). Of course I think these probabilities probably overstate Duke and UVa's real chances since they are based on performance to-date and therefore don't account for things like potential injuries, etc., but maybe this gives an idea of the magnitude of the impact of schedule differences.

For my edification, do these all start with some sort of current ranking of opponents?

[vick]

For my edification, do these all start with some sort of current ranking of opponents?

Yes, they're calculated based on the current Kenpom offensive and defensive efficiencies of each team and their opponents, i.e., I use the efficiencies to estimate the chance of winning on a per-game basis (using the pythagorean formula here (http://kenpom.com/blog/index.php/weblog/entry/ratings_explanation)), then use Excel's random number generator (which generates a random number between 0 and 1) and compare. So if Duke has a 75% chance of winning a game, and the generator spits out anything from 0 to 0.75, this is a simulated Duke win. I just looped this repeatedly to simulate a few thousand seasons.

[CDu]

I agree that his "luck" is actually a mix of luck and model flaws to some degree. I do think it's possible to get some sort of idea on how much of each is present. For example: my guess is that end of game situations involving intentional fouling don't work very well in a strict pace-adjusted system, which would cause slow teams that are good at free throw shooting (like Wisconsin and Virginia) to be slightly overrated. This is something that should be relatively simple to test.

Another example is that I don't think it's unreasonable to suggest that some teams are genuinely better at winning close games. Even if you don't buy into clutch shooting (which I don't), you can still have effects like teams insisting that Kobe Bryant always take the last shot in a one possession game, which decreases the chances of winning while having only a very minor effect on efficiency numbers. Which is admittedly a little harder to test for 351 teams.

The challenge is how would you control for that? Intentional fouling starts at different times each game. Furthermore, there is also the possibility of improved opponents' offensive efficiency as a result of sort of a prevent defense (contest threes but allow twos and avoid fouling) leading to easier buckets. And if you consider late-game fouling, what about stallball? It is just really really complicated, and we don't have a huge sample size to work with. Hence my point about sample size and large number of variables.

[sagegrouse]

This, unfortunately, is beyond the skills that I dimly remember from Stat 103 a decade ago, but I think I was able to successfully simulate 10,000 seasons with the real schedule and a 28-game double round robin schedule, and here are the chances of winning the conference title (I removed teams with very low chances for the sake of readability):

For the sake of the clueless, what do you do to estimate the chance of one team winning a single game? does KenPom or someone provide a "win probability" formula based on current ratings?

[uh_no]

For the sake of the clueless, what do you do to estimate the chance of one team winning a single game? does KenPom or someone provide a "win probability" formula based on current ratings?

kenpom gives those numbers on the team page. I know how to come up with a rough estimate (teamA.offense + teamB.defense - average.offense for each set of A and B teams, then pythagorean to get a win probability...which is offense^x / (offense^x + defense^x) where he uses x=10.25), which seems to pretty closely match his estimates.

Someone may know if he lists the formula anywhere

EDIT: http://207.56.97.150/articles/playoff2002.htm that's the formula he linked in 2006 from http://kenpom.com/blog/index.php/weblog/entry/ratings_explanation

[vick]

For the sake of the clueless, what do you do to estimate the chance of one team winning a single game? does KenPom or someone provide a "win probability" formula based on current ratings?

It's very close to what uh_no describes below. I use (and I get percentages that match Kenpom except for the rare rounding error):

Expected Game Offense Efficiency = (League Average)*(Off Efficiency/League Average)*(Opponent Def Eff/League Average)
Expected Game Defense Efficiency = (League Average)*(Opponent Off Efficiency/League Average)*(Def Efficiency/League Average)

That's questionable algebra since it simplifies a bit, but it's easier to think of it this way: offense = average times the percentage your offense is above average times the percentage the opponent's defense is above average (where a lower number is better for defense).

I add a home-court advantage of 1.4% to both offense and defense, and then use the formula uh_no posted.

[uh_no]

It's very close to what uh_no describes below. I use (and I get percentages that match Kenpom except for the rare rounding error):

Expected Game Offense Efficiency = (League Average)*(Off Efficiency/League Average)*(Opponent Def Eff/League Average)
Expected Game Defense Efficiency = (League Average)*(Opponent Off Efficiency/League Average)*(Def Efficiency/League Average)

That's questionable algebra since it simplifies a bit, but it's easier to think of it this way: offense = average times the percentage your offense is above average times the percentage the opponent's defense is above average (where a lower number is better for defense).

I add a home-court advantage of 1.4% to both offense and defense, and then use the formula uh_no posted.

Thanks for that! it reduces to A.off * B.def / average.off, which is the geometric version of the formula i was using. Over the range where both teams are between 80 and 120 (which encompasses every team except duke's offense at just over 120), the error except in the extreme cases is less than 1 point per 100 posessions. The errors are:

great offense and great defense (or poor offense and poor defense): simple model will be too generous to the offense
great offense vs poor defense (or poor offense vs great defense): simple model will be too generous to the defense

Even in these extreme circumstance, the total error is is less than 4pphp in either direction...so it's a decent estimate.

Intuitively, simple model says "for every 10 points better you are in offense, you will be 10 points better versus ANY team." Complex model says "Against an 80 defense, every 10 points of offense only nets you EIGHT points, whereas versus a 110 defense, every 10 points of offense nets you ELEVEN points."

Here's a nice table:
Team with given efficiency playing a team with defensive efficiency of 110:



100
110
120


simple vs 110
110
120
130


cmplx vs 110
110
121
132


simple vs 80
80
90
100


cmplx vs 80
80
88
96



So that gives you a relatively easy way to estimate...for every 10 you are above 100 on offense, add a tenth of your opponents defense to your opponents defense. Here's how it would work with, say, duke vs UK:
duke is a 121.4, UK is an 81.7....so if duke were a 120...we'd add 1/10 of UK's defense twice:

81.7+8.1+8.1 ~= 98 or 99 (since duke is a bit above 120)

Yes it's a stupid math trick, but also an easy way to think about it.

[JasonEvans]

That was the most dominant Duke basketball team I can remember. Against Cincinnati, I believe we simply ran out of time...Will Avery had either a tying shot or a game-winning shot before overtime that was waived off because the clock expired. The <10 point victories were 2x against next year's champion Michigan State (6 points each time), @Ga Tech was an 8 point victory, and the St John's game.

Regarding undefeated, Duke was only half a second late against Cincinnati, and a Trajan Langdon travel call against UConn away.

I also remember JJ Redick's senior season, we started out something like 17-0 or 18-0, and lost at Georgetown. That was the year that we blew out Texas in MSG.

1999 is the gold standard for a regular season IMHO, and this year is the closest I can remember to feeling that way. Obviously we have a long way to go.

I think you may not be recalling what 1999 was like. This year's team is playing great, but they ain't the 1999 Blue Devils, at least not yet. By this point in the 98-99 season (early January) the Langdon/Brand/Avery/CWell/Burgess/Battier/Maggette bunch had been blowing the doors off of good teams. The beat Notre Dame by 29, Florida by 30, and Michigan by 44. #3 Kentucky went down by 11. They began January by stomping #4 Maryland by 18... at Maryland. They following that game by beating Ga Tech by 41 and Virginia by 46. Most ACC games were over after ten or fifteen minutes. Being in a game with Duke at halftime was a great accomplishment.

True, 1998-99 had a loss at this point, a big strike versus the unblemished record of this year's club, but 1999 felt more like an unstoppable juggernaut to me. Of course, given how 1999 ended, maybe being great, but not quite an untouchable juggernaut is an ok thing ;)

-Jason "one difference to me is the presence of Kentucky -- in 1999, we were pretty clearly the #1 team, while this year - even undefeated -- everyone sees us as the #2" Evans

[Wander]

-Jason "one difference to me is the presence of Kentucky -- in 1999, we were pretty clearly the #1 team, while this year - even undefeated -- everyone sees us as the #2" Evans

To be fair, I think that says more about Kentucky than about us. The only recentish year I can think of where this team wouldn't be the obvious #1 ranked team at this date in the season is 2006 (and the team in question from that year is also us!)

[Mtn.Devil.91.92.01.10.15]

I think you may not be recalling what 1999 was like. This year's team is playing great, but they ain't the 1999 Blue Devils, at least not yet. By this point in the 98-99 season (early January) the Langdon/Brand/Avery/CWell/Burgess/Battier/Maggette bunch had been blowing the doors off of good teams. The beat Notre Dame by 29, Florida by 30, and Michigan by 44. #3 Kentucky went down by 11. They began January by stomping #4 Maryland by 18... at Maryland. They following that game by beating Ga Tech by 41 and Virginia by 46. Most ACC games were over after ten or fifteen minutes. Being in a game with Duke at halftime was a great accomplishment.

True, 1998-99 had a loss at this point, a big strike versus the unblemished record of this year's club, but 1999 felt more like an unstoppable juggernaut to me. Of course, given how 1999 ended, maybe being great, but not quite an untouchable juggernaut is an ok thing ;)

-Jason "one difference to me is the presence of Kentucky -- in 1999, we were pretty clearly the #1 team, while this year - even undefeated -- everyone sees us as the #2" Evans

To be fair, you might be making his point for him - that the 1999 team is the far-and-away standard for best regular season. And, that this year (thus far) may still be the closest thing, even if there's a wide gap. What season would be the better candidate for second place?

[AIRFORCEDUKIE]

I think you may not be recalling what 1999 was like. This year's team is playing great, but they ain't the 1999 Blue Devils, at least not yet. By this point in the 98-99 season (early January) the Langdon/Brand/Avery/CWell/Burgess/Battier/Maggette bunch had been blowing the doors off of good teams. The beat Notre Dame by 29, Florida by 30, and Michigan by 44. #3 Kentucky went down by 11. They began January by stomping #4 Maryland by 18... at Maryland. They following that game by beating Ga Tech by 41 and Virginia by 46. Most ACC games were over after ten or fifteen minutes. Being in a game with Duke at halftime was a great accomplishment.

True, 1998-99 had a loss at this point, a big strike versus the unblemished record of this year's club, but 1999 felt more like an unstoppable juggernaut to me. Of course, given how 1999 ended, maybe being great, but not quite an untouchable juggernaut is an ok thing ;)

-Jason "one difference to me is the presence of Kentucky -- in 1999, we were pretty clearly the #1 team, while this year - even undefeated -- everyone sees us as the #2" Evans

Except for Len Elmore, he sees us as #4 or #5 ;)

[Des Esseintes]

To be fair, you might be making his point for him - that the 1999 team is the far-and-away standard for best regular season. And, that this year (thus far) may still be the closest thing, even if there's a wide gap. What season would be the better candidate for second place?

I think 2001-02 gets neglected because we lost in the Sweet 16 and Maryland ended up winning the title. Hangover from the title year probably had an effect, too; dominant as we were, it was hard not to wish we were even more dominant. But man, that team was good. #1 in both offense and defensive efficiency on kenpom. Like this year's team, undefeated through December. After losing to Florida State on the road, Duke just blitzed the league for six weeks. 11 straight double-digit victories, seven by margins greater than 20 points. It blew out eventual national champion Maryland by 21 in Cameron. Duke swept Carolina in three games by a combined 66(!) points. ACC championship. 4th straight year getting a 1-seed going into the NCAA tournament. 4th straight year ending the regular season ranked #1. There was just a lot to enjoy from that season. This year's team would have to do some stuff to equal it.

[-jk]

I think you may not be recalling what 1999 was like. This year's team is playing great, but they ain't the 1999 Blue Devils, at least not yet. By this point in the 98-99 season (early January) the Langdon/Brand/Avery/CWell/Burgess/Battier/Maggette bunch had been blowing the doors off of good teams. The beat Notre Dame by 29, Florida by 30, and Michigan by 44. #3 Kentucky went down by 11. They began January by stomping #4 Maryland by 18... at Maryland. They following that game by beating Ga Tech by 41 and Virginia by 46. Most ACC games were over after ten or fifteen minutes. Being in a game with Duke at halftime was a great accomplishment.

True, 1998-99 had a loss at this point, a big strike versus the unblemished record of this year's club, but 1999 felt more like an unstoppable juggernaut to me. Of course, given how 1999 ended, maybe being great, but not quite an untouchable juggernaut is an ok thing ;)

-Jason "one difference to me is the presence of Kentucky -- in 1999, we were pretty clearly the #1 team, while this year - even undefeated -- everyone sees us as the #2" Evans

Jason, I think you're misrepresenting 1999. While Duke was blowing doors off, we hadn't been #1 since the end of November, and wouldn't regain #1 until the second week of February. UConn was #1 in the interim. Duke and UConn were the only #1 teams all season, and both Duke and UConn were top 5 all season.

-jk

[sagegrouse]

To be fair, you might be making his point for him - that the 1999 team is the far-and-away standard for best regular season. And, that this year (thus far) may still be the closest thing, even if there's a wide gap. What season would be the better candidate for second place?

The 1992 season, no question, and the team was clearly head and shoulders above the other NCAA teams. Only two losses -- at UNC by two and at Wake by four while Hurley was out. While the margin of victory was less gaudy than 1999 -- 15.4 vs. 24.6 -- the 1992 team didn't run up the score. Instructive was an early season match-up at Cameron against a tough St. John's team, which Duke won by ten. Actually, Duke was ahead by 30 and then just eased up noticeably on the throttle. The 1992 team was defending national champions, knew it was the best team and toyed with the opposition.

[Kedsy]

...in 1999, we were pretty clearly the #1 team, while this year - even undefeated -- everyone sees us as the #2

Well, except on January 5, 1999, Duke was #2, behind Connecticut, which had been #1 for all except the first couple weeks that season.

[throatybeard]

Y'all are confusing me with the probability jazz.

So if there are three doors, and Kaylee Hartung is behind one of them, and Coach K opens one door and shows me a goat, should I switch my original choice of door? Please advise.

[Kfanarmy]

Y'all are confusing me with the probability jazz.

So if there are three doors, and Kaylee Hartung is behind one of them, and Coach K opens one door and shows me a goat, should I switch my original choice of door? Please advise.

That depends on the probability that your spouse is standing behind you.

[OldPhiKap]

Y'all are confusing me with the probability jazz.

So if there are three doors, and Kaylee Hartung is behind one of them, and Coach K opens one door and shows me a goat, should I switch my original choice of door? Please advise.

Depends on preference and proclivities, not probability.

[throatybeard]

Depends on preference and proclivities, not probability.

http://imgs.xkcd.com/comics/monty_hall.png

[Olympic Fan]

Just looked at the last three unbeaten teams on Pomeroy.

He now has Kentucky favored in every remaining game and given almost a 30 percent chance to finish (the regular season) unbeaten.

But he also Has Virginia favored in every remaining game (64 percent chance of beating Duke in C'ville) ... and a 5.5 percent chance of finishing unbeaten.

He has Duke favored in ever other game (just flipped the Duke at Louisville game ... Duke is now a 52 percent favorite. He has Duke as a 1.5 percent chance of finishing unbeaten.

Jason's point about the 1999 team is interesting, but a little misleading. At this point in 1999, Duke had lost a game and won a six-point game from No. 9 Michigan State on a neutral court (Mich State would reach the Final Four, where Duke would beat them by six again).

But a couple of the teams you cite were pretty mediocre -- Notre Dame finished 14-16 that season and Michigan was 12-19 (they later vacated those 12 wins, so officially they finished 0-19). The Georgia Tech and Virginia wins you cite came after Jan. 5.

At this point in 1999, Duke had a close lose to a very good Cincinnati team (one that would finished No. 11 in the final AP poll), a blowout win at home over a good Florida team (22-9 and No. 22 in the final poll) and a good, hard-fought win in the Meadowlands over Kentucky (No. 11 in the final polls) -- it ended up 11 points, but Duke didn't put the game away until late ... very much like his year's 10-point win an Wisconsin. The Jan. 3 victory over Maryland in College Park was the team's most impressive win -- 18 points over the team that would finish No. 5 in the AP poll.

I was the one who suggested that the 1999 team was Duke's greatest regular season squad. But they really took off as an absolutely dominant team about this time. If you compare 1999 Duke to 2015 Duke AS OF JAN, 5, I think it's a close call. You can make a reasonable argument that this team is off to a better start. The '99 team had two great wins (Kentucky and Maryland), plus a close win over Michigan State ... but they also had a loss. The '15 team doesn't have that many impressive wins (although at No. 2 Wisconsin is as good as anything '99 did to this point), but it also doesn't have a loss.

Right now, it's close -- but if 2015 wants to keep up with 1999, it will have to really turn on the jets.

[Henderson]

Just looked at the last three unbeaten teams on Pomeroy.

He now has Kentucky favored in every remaining game and given almost a 30 percent chance to finish (the regular season) unbeaten.

But he also Has Virginia favored in every remaining game (64 percent chance of beating Duke in C'ville) ... and a 5.5 percent chance of finishing unbeaten.

He has Duke favored in ever other game (just flipped the Duke at Louisville game ... Duke is now a 52 percent favorite. He has Duke as a 1.5 percent chance of finishing unbeaten.

I don't know why teams bother to play games if Pomeroy has already decided the outcomes. Think of the travel money teams would save.

Sorry, I didn't mean to disparage anyone's religion.

[JasonEvans]

I don't know why teams bother to play games if Pomeroy has already decided the outcomes. Think of the travel money teams would save.

Sorry, I didn't mean to disparage anyone's religion.

Are you being intentionally obtuse or are you unclear on the difference in probability and outcome?

-Jason "if you locked Ken Pomeroy and Nate Silver in a room for a week, what are the odds they could come up with the winning lottery numbers? Ha!" Evans

[throatybeard]

I don't know why teams bother to play games if Pomeroy has already decided the outcomes. Think of the travel money teams would save.

Sorry, I didn't mean to disparage anyone's religion.

Maybe he can use his powers for good and put an end to these idiotic NFL in London games.

[Henderson]

Are you being intentionally obtuse or are you unclear on the difference in probability and outcome?


Two things can be distinct and yet inextricable. The only reason that probabilies in this context are useful is in relation to the forecasting of outcomes.

Well, that and giving folks something to talk about. :rolleyes:

[Mtn.Devil.91.92.01.10.15]

Well, I would rather Duke be the last team to get a victory at the end of the season than the last team with a loss at the beginning...

[ACCBBallFan]

Whille only UVA and UK are contenders for Unbeaten, there are also only 3 teams with only one loss now that both Zona and Wisc were also beaten

Team Conference W-L Pomeroy

Kentucky SEC 15-0 0.9765
Virginia ACC 15-0 0.9757

Gonzaga WCC 16-1 0.9534
Duke ACC 14-1 0.9515
Villanova BE 15-1 0.9506

[MCFinARL]

And now there is one. :D

[CDu]

Two things can be distinct and yet inextricable. The only reason that probabilies in this context are useful is in relation to the forecasting of outcomes.

Well, that and giving folks something to talk about. :rolleyes:

There is a difference between forecasting the outcome and deciding the outcome...

[Olympic Fan]

At the risk of upsetting Henderson, I should note that Pomeroy now has Kentucky rated as a 49.0 percent chance of finishing the regular season unbeaten.

Toughest games left? At Florida (86 percent chance of a Kentucky win) and at Georgia (86 percent chance).

No more ranked opponents, which is obvious since the SEC doesn't have a ranked team other than Kentucky.