MBB Dork Polls/Stats: 2017-18 Edition

[Troublemaker]

Kenpom's pre-season rankings are up: https://kenpom.com/

Duke checks in at #6 overall, with the projected #1 offense and #35 defense.

Not too surprising that we would be ranked so "low" since we finished #14 and #17 overall the past two seasons. We also have a lot to prove defensively.

[UrinalCake]

If we were actually the top offense and 35th best defense in the country right now, stating four freshmen and a senior with two sophomores coming off the bench, I'd be pretty thrilled with that. We're only going to improve as the season progresses.

[uh_no]

If we were actually the top offense and 35th best defense in the country right now, stating four freshmen and a senior with two sophomores coming off the bench, I'd be pretty thrilled with that. We're only going to improve as the season progresses.

Not sure evidence supports that.

We had one year in the one and done era in which the defense came together down the stretch. Unfortunately we don't have daily data for the past 7 years, but duke has typically started pretty highly, and often has dropped significantly as the year goes on.

Now, (again from memory), we usually pull that number back somewhat from the low points of the year, but I don't believe it ever really reaches the preseason expectation.

[rsvman]

How does Ken even do rankings for offense and defense before a single game has been played?

[CDu]

How does Ken even do rankings for offense and defense before a single game has been played?

It is basically meaningless. But he does it based on a combination of returning player contributions, incoming freshmen (based on historical production of freshmen), and the program’s recent performance.

But this year’s team really doesn’t work for any type of preseason statistical analysis.

[sagegrouse]

How does Ken even do rankings for offense and defense before a single game has been played?

His "method" actually works only after a sizable number of games have been played -- as in, like, January. But he has subscribers who are customers, and customers want estimates for the entire season. So, as I understand it (I heven't pored through the details), he has come up with a preseason measure based on last season, returning players, new players, and a large dose of heuristics.* Then he gradually reduces the "preseason estimate" as he gets more data from 2017-2018 games. By early-to-mid-January, his data-driven estimates are 100 percent.

Kindly,
Sage Grouse

*Judgment

[COYS]

Not sure evidence supports that.

We had one year in the one and done era in which the defense came together down the stretch. Unfortunately we don't have daily data for the past 7 years, but duke has typically started pretty highly, and often has dropped significantly as the year goes on.

Now, (again from memory), we usually pull that number back somewhat from the low points of the year, but I don't believe it ever really reaches the preseason expectation.

I know the injury vs. OAD debate has been litigated a number of times, but I think this is a generalization that requires a little bit of nuance. KenPom does not factor injuries into his numbers. Duke has had poor (relatively speaking) defensive numbers in four out of the last seven seasons so 35 sounds about right to me in a vacuum. However, Duke also had major injury issues in three of those four bad seasons.

It's not surprising Duke's overall rating would more often decrease because that is the more likely for a team ranked near the top. Last year, Kenpom started Duke number one, but all that meant is that Duke had, per his numbers, the best percentage chance of being number one in the rankings. I'd bet that the actual chance Duke had of being number one was really, really small in isolation. (Kentucky 2x, Louisville 2x, and Duke 3x are the only teams in the KenPom era with multiple number one rankings, so Duke is 3/17 over that span, good for 18% which actually sounds pretty high to me) So any team ranked at the top has a MUCH better chance of falling lower than remaining on top.

Then, of course, injuries have also plagued Duke teams. Duke didn't have any major players miss significant time save for Carlos Boozer from 1999-2010, which was pretty phenomenal. Yes, some guys had some ailments, obviously, but there were no extended absences, especially from the teams' stars. That trend reversed in a big way starting with 2011 with Kyrie, 2012 and 13 with Ryan, 2016 with Amile, and 2017 with basically everyone. I definitely think there is strong evidence that veteran teams were able to handle big losses better than younger teams. 2011 was capable of winning the title with or without Kyrie. 2013 hung in there during Ryan's absence and was also a legitimate contender at the end of the season. Meanwhile, 2012 fell apart after Ryan's injury. 2016 played well on offense but never got it together on defense after Amile's injury. Meanwhile, despite the disappointing finish, the veteran-laden rotation for the 2017 team actually did pretty well considering all the injuries and drama, recovering to win the ACCT.

The team that relied on OAD's the most was 2015, which also happens to be the team that managed to improve the most on D by the end of the season. And I think the 2017 team was hurt by NOT being able to rely more on a healthy and freaky-athletic Harry Giles . . . so one could actually argue that 2017 ended up worse because we weren't able to rely on a OAD star more.

This isn't to say that there isn't good evidence that relying on OAD's makes for a wider variance in results. I would argue that the reasons for this have less to do with having any particular OAD on the roster and more to do with the overall higher turnover in personnel between years. That is definitely a concern for this coming season. But I think there is reason to be optimistic that if Duke stays healthy, Duke will be good, even with OAD talent.

[MarkD83]

Do exhibition games count in the Kenpom stats

[Kedsy]

Do exhibition games count in the Kenpom stats

No, they don't.

[Wander]

Duke has had poor (relatively speaking) defensive numbers in four out of the last seven seasons so 35 sounds about right to me in a vacuum. However, Duke also had major injury issues in three of those four bad seasons.


I don't think our poor defensive numbers can be blamed on injuries in any of those seasons, except for last year.

[Kedsy]

I don't think our poor defensive numbers can be blamed on injuries in any of those seasons, except for last year.

I'd add 2013 and 2016. In 2013, we had a top 10 defense (per Pomeroy) when Ryan Kelly got hurt. In 2016, hard to imagine our D not being better if Amile had played all season.

[Wander]

I'd add 2013 and 2016. In 2013, we had a top 10 defense (per Pomeroy) when Ryan Kelly got hurt. In 2016, hard to imagine our D not being better if Amile had played all season.

I wouldn't say 2013 was a poor defensive season to start with, so that year shouldn't be included.

[slower]

I'd add 2013 and 2016. In 2013, we had a top 10 defense (per Pomeroy) when Ryan Kelly got hurt. In 2016, hard to imagine our D not being better if Amile had played all season.

Perhaps significantly so.

[Olympic Fan]

Okay, I know that there are much better statistical measures of a season that RPI.

But the RPI remains the selection committee's favorite tool, so we need to pay attention to it.

I also know that it's too early to pay much attention to RPI rankings (or anything else, such as Ken Pom). Still, just for fun, I checked out the RPI as of today:

https://www.teamrankings.com/ncb/rpi/

Duke is 19 today ... and it's not likely to get a lot better until ACC play starts,

Duke's upcoming opponents are not all that -- Portland State is 339, Butler is 106, Texas is 98, Florida is No. 7 in the nation in the AP and No. 248 in RPI. Gonzaga is 148. Indiana is 203.

I take those rankings of evidence as to how the early season RPI can be warped. Still, it will be interesting to watch the rankings evolve as we get deeper into the season.

[uh_no]

should this be wrapped up in a dork polls thread?

In kenpom news, arizona's demise bumped us up.

[UrinalCake]

I expect our strength of schedule at the end of the season to be pretty bad, and in turn our RPI. Weak OOC schedule outside of MSU, and a favorable conference schedule. Which is good in the sense that we’ll take fewer losses, but bad in terms of computer numbers.

[Troublemaker]

I expect our strength of schedule at the end of the season to be pretty bad, and in turn our RPI. Weak OOC schedule outside of MSU, and a favorable conference schedule. Which is good in the sense that we’ll take fewer losses, but bad in terms of computer numbers.

RPI Forecast uses Sagarin to project the season, and Duke ends up as #3 RPI and #17 SOS (http://www.rpiforecast.com/index2.html). Not bad at all. We have enough tough games on the schedule. As mentioned elsewhere, I don't think the ACC will suffer through a down year.

[arnie]

RPI Forecast uses Sagarin to project the season, and Duke ends up as #3 RPI and #17 SOS (http://www.rpiforecast.com/index2.html). Not bad at all. We have enough tough games on the schedule. As mentioned elsewhere, I don't think the ACC will suffer through a down year.

Yea, these projections have the esteemed UNC program at #2. Don’t think that level even remotely holds as the season progresses.

[uh_no]

Yea, these projections have the esteemed UNC program at #2. Don’t think that level even remotely holds as the season progresses.

what is sagarin's pre-season numbers? If it's anything like KP, UNC gets a huge bump due to really really good performance the past couple of years

[ChillinDuke]

Interesting tidbits to me...

KP lists Grayson Allen with a sensational oRTG of 139.7, good for 17th in the nation among all players, not just qualifiers. What is of further interest to me is that KP still lists G as a "Role Player" at 16-20% possession usage. For reference, Trae Young is #1 in the country in oRTG at 130.3 for all players with at least 28% usage. Not coincidentally, KP lists Trae as his early frontrunner for NPOY.

Pivoting for a second to Trae Young, he has been exceptional this year, averaging 28.8 and 8.8 assists, dropping 20+ in 5 of his 6 games. But I wonder if Oklahoma will keep pace in the Big 12 enough to keep Trae up near the top of the NPOY race. Similar issue to Luke Maye, sort of.

On this young season, my NPOY shortlist is Bagley, Devonte Graham, Colson, Bridges on Villanova, and probably this Murphy guy on Minnesota, all of which are listed on KP's listing. If Trae Young and Luke Maye can keep it up, then they'll be in the mix - I just don't think they will, at least at this level. The other guys I think will.

- Chillin

[CDu]

Interesting tidbits to me...

KP lists Grayson Allen with a sensational oRTG of 139.7, good for 17th in the nation among all players, not just qualifiers. What is of further interest to me is that KP still lists G as a "Role Player" at 16-20% possession usage. For reference, Trae Young is #1 in the country in oRTG at 130.3 for all players with at least 28% usage. Not coincidentally, KP lists Trae as his early frontrunner for NPOY.

Pivoting for a second to Trae Young, he has been exceptional this year, averaging 28.8 and 8.8 assists, dropping 20+ in 5 of his 6 games. But I wonder if Oklahoma will keep pace in the Big 12 enough to keep Trae up near the top of the NPOY race. Similar issue to Luke Maye, sort of.

On this young season, my NPOY shortlist is Bagley, Devonte Graham, Colson, Bridges on Villanova, and probably this Murphy guy on Minnesota, all of which are listed on KP's listing. If Trae Young and Luke Maye can keep it up, then they'll be in the mix - I just don't think they will, at least at this level. The other guys I think will.

- Chillin

Yeah, the "role player" designation is just a shorthand for % usage category. We are probably not the best team to use that shorthand, as we have been a really balanced offense this year, with 3-4 guys hovering right around 20% usage (falling in the "role player" category) and one comfortably above it. Whereas most teams have 2-3 guys who dominate the possessions, we just don't. We have one superstar and 4 other stars, and those 4 other stars share the possessions not used by the superstar.

[proelitedota]

I wrote to Kenpom regarding an observation that I made after the South Dakota game. He responded pretty quickly.


Greetings!

Love you site as always but there is one aspect of your algo that can be improved.

I.e. Team is up 30 with 10 minutes to ago against an overmatched team in which the spread was 35, starters are pulled and the other team pulls within 15 in the final margin. Currently it seems as your algorithm is penalizing the stronger team for pulling the starters and letting the opponent hang around in final meaningless minutes.

Keeping it short!



I agree, that is a shortcoming of the system.


I think he can improve it if his system can track the substitution patterns.

[uh_no]

I wrote to Kenpom regarding an observation that I made after the South Dakota game. He responded pretty quickly.





I think he can improve it if his system can track the substitution patterns.

ooooooo

now the only question is if his system is intelligent enough to respond to fanmail :D :D

cool.

I think it's a difficult problem to solve generally. How do you determine which possessions in a game "count"? When 1 starter is out? 2? 4? What if we play some minor guys in the first half (like we have several times this year already)...do those minutes count?

While there may be some ways to improve it slightly, I think at the end of the year, those 5 minutes are going to mean so little in the grand scheme of things (in a game that will already be pretty trivialized due to how long ago it was and how it didn't deviate a huge amount from the "blowout" prediction)

[BandAlum83]

ooooooo

now the only question is if his system is intelligent enough to respond to fanmail :D :D

cool.

I think it's a difficult problem to solve generally. How do you determine which possessions in a game "count"? When 1 starter is out? 2? 4? What if we play some minor guys in the first half (like we have several times this year already)...do those minutes count?

While there may be some ways to improve it slightly, I think at the end of the year, those 5 minutes are going to mean so little in the grand scheme of things (in a game that will already be pretty trivialized due to how long ago it was and how it didn't deviate a huge amount from the "blowout" prediction)

I do think it's pretty clear that the only way we get to 100 points in a game is by playing a really good opponent in a competitive game. We pull off the accelerator against lesser teams to such an extent, it almost seems Coach K doesn't really want the team showing up an opponent to that extent.

Maybe the annual predictions podcast shouldn't even have that as a category any longer.

Now that I've commented on this, we are sure to score 108 on Tuesday. :)

[uh_no]

I do think it's pretty clear that the only way we get to 100 points in a game is by playing a really good opponent in a competitive game. We pull off the accelerator against lesser teams to such an extent, it almost seems Coach K doesn't really want the team showing up an opponent to that extent.

Maybe the annual predictions podcast shouldn't even have that as a category any longer.

Now that I've commented on this, we are sure to score 108 on Tuesday. :)

I think K would prefer 100 points to what the end of the bench guys did the other night. You don't need to show off and run up the score to be playing well....and I'd like to think K wants every guy on the floor to play well.

[House P]

ooooooo

now the only question is if his system is intelligent enough to respond to fanmail :D :D

cool.

I think it's a difficult problem to solve generally. How do you determine which possessions in a game "count"? When 1 starter is out? 2? 4? What if we play some minor guys in the first half (like we have several times this year already)...do those minutes count?

While there may be some ways to improve it slightly, I think at the end of the year, those 5 minutes are going to mean so little in the grand scheme of things (in a game that will already be pretty trivialized due to how long ago it was and how it didn't deviate a huge amount from the "blowout" prediction)

One way to tweak the system would be to reduce the weighting of possessions which occur once the outcome of a game is no longer in doubt. For example, you could reduce the weight of any possession which occurred after a team has established a 99.5% chance of winning the game. This would certainly complicate the calculations (which would now have to be done on a play-by-play basis instead of just looking at the final boxscore), but would probably improve things at least a tiny bit.

I suspect that this may be a factor which occasionally causes extremely low tempo teams (UVA, Wisconsin) to appear over-rated KenPom's rankings. If a game only has 55 possessions, it is unlikely that one team will build a 30+ point lead midway through the second half, empty the bench, and get outscored over the final 5-10 minutes of the game. And even if this occurs, my understanding is that a 20 point win in a 55 possession game still gets you a better rating than a 20 point win in a 75 possession game.

[uh_no]

One way to tweak the system would be to reduce the weighting of possessions which occur once the outcome of a game is no longer in doubt. For example, you could reduce the weight of any possession which occurred after a team has established a 99.5% chance of winning the game. This would certainly complicate the calculations (which would now have to be done on a play-by-play basis instead of just looking at the final boxscore), but would probably improve things at least a tiny bit.

I suspect that this may be a factor which occasionally causes extremely low tempo teams (UVA, Wisconsin) to appear over-rated KenPom's rankings. If a game only has 55 possessions, it is unlikely that one team will build a 30+ point lead midway through the second half, empty the bench, and get outscored over the final 5-10 minutes of the game. And even if this occurs, my understanding is that a 20 point win in a 55 possession game still gets you a better rating than a 20 point win in a 75 possession game.

In a low tempo game, however, the win percentage for a 10 point lead should be higher than it is in a high tempo game. should....in theory....but i don't think those live projections are so fancy.

Another way, if you're up 15 in a slow game, your changes of winning might be the same as being up 30 in a fast game....but then how do you account if the game speeds up?

Either way, i'd be skeptical of using TOO much of by-posession win % in the ultimate rankings....it would mean those % would have to be near the robustness of his overall algorithm...and if they are not, it may have a detrimental impact on the final rankings.

in his last blog post, he talked about how the win % don't necessarily agree with his main system's predicted win %...even though they are usually close.

Its a nice idea, but as you point out, there are some potential pitfalls that would require a lot of thought to get right.

[toooskies]

"Fixing" late-game blowouts can be a logical progression of the work that KenPom does with in-game percent chances to win. You set some scale based on % chance to win (i.e. only count possessions when <99% and >1% chance to win; scale possession value based on difference between a successful scoring possession or not; or whatever other scale you like) and multiply out the value of possessions from there. The choice of scale depends on whether you want to be more predictive or descriptive: being "clutch" often isn't repeatable, but it's still notable.

[uh_no]

"Fixing" late-game blowouts can be a logical progression of the work that KenPom does with in-game percent chances to win. You set some scale based on % chance to win (i.e. only count possessions when <99% and >1% chance to win; scale possession value based on difference between a successful scoring possession or not; or whatever other scale you like) and multiply out the value of possessions from there. The choice of scale depends on whether you want to be more predictive or descriptive: being "clutch" often isn't repeatable, but it's still notable.

(see above couple of posts)

It's not a bad idea, but it's not trivial to do in a clean way.

[Kedsy]

I suspect that this [tempo] may be a factor which occasionally causes extremely low tempo teams (UVA, Wisconsin) to appear over-rated KenPom's rankings. If a game only has 55 possessions, it is unlikely that one team will build a 30+ point lead midway through the second half, empty the bench, and get outscored over the final 5-10 minutes of the game. And even if this occurs, my understanding is that a 20 point win in a 55 possession game still gets you a better rating than a 20 point win in a 75 possession game.

I know KenPom has been trying to fix the "Wisconsin problem" for some time, and I know one popular theory is that it's related to slow tempo, so I examined all 16 years of Pomeroy data to see if teams with slow tempo fare worse in the NCAA tournament than we'd expect.

At the outset, I admit there are lots of potential flaws in this analysis. First, what's called a "slow" tempo is arbitrary. For my population of teams, I chose KenPom top 20 teams with a tempo ranked 240 or worse. Second, NCAA tournament success as a measure of overrated or underrated is probably wrong, because of the arbitrary nature of the Tournament. Although wrong or not, most people seem to look at it that way. Third, I'm measuring expected performance assuming all higher-seeded teams should win, and actual performance based on the same measurement. For example, I'm expecting a 3-seed to make the Sweet 16 then lose (to the 2-seed), and if that's what a 3-seed does I deem it as meeting expectations, even if the 3-seed lost to a 10-seed in its Sweet 16 game. Fourth, I used post-tournament data, which is sort of a self-fulfilling prophecy, but I didn't have a full complement of pre-tourney data. And fifth, I'm using Pomeroy's most-recent, "Wisconsin-proof" rating system, which may (or may not) make the whole thing moot. Sixth, the better a team is seeded, the harder it is to overachieve and the easier it is to underachieve; the worse a team is seeded, the opposite is true.

There have been 91 "slow" KenPom top 20 teams that played in the tournament since 2001-02 (earliest year for which KenPom posts results). Presumably because of their slow tempo, most of these teams (47, or 51.6%, were underseeded, based on their KenPom rank). Only 16 of these teams (17.6%) were overseeded, and 28 of them (30.8%) were seeded properly. Note that of the 16 "overseeded" teams, only six of them met or exceeded expectations.

"Slow" KenPom top 20 teams seeded 9 or worse ("expected" to win zero games):

6 teams; all six won at least one game (100%); 2 of 6 won multiple games (33.3%).

Historically (since the tourney went to 64 teams), 41.7% of all teams seeded 9 to 11 won one game, and 12.1% won multiple games.


"Slow" KenPom top 20 teams seeded 5 to 8 ("expected" to win one game):

34 teams; 25 won at least one game (73.5%); 16 won multiple games (47.1%).

Historically, 59.8% of teams seeded 5 to 8 won at least one game, and 31.1% won multiple games.


"Slow" KenPom top 20 teams seeded 3 or 4 ("expected" to win two games):

20 teams; 14 won two or more games (70.0%); 6 were upset early (30.0%).

Three of the six upset victims were overseeded, based on KenPom rank; if you don't count overseeded teams, we have 13 teams and 11 of them (78.6%) won two or more games.

Historically, only 49.2% of teams seeded 3 or 4 win two or more games, and 50.8% are upset early.


"Slow" KenPom top 20 teams seeded 2 ("expected" to win three games):

19 teams; 8 won three or more games (42.1%); 11 were upset early (57.9%).

Five of the eleven upset victims were overseeded, based on KenPom rank; if you don't count overseeded teams, we have 13 teams and 6 of them (46.2%) won three or more games.

Historically, 47.7% of 2-seeds won three or more games, and 52.3% were upset early.


"Slow" KenPom top 20 teams seeded 1 ("expected" to win four games):

12 teams; 4 won four or more games (33.3%); 8 were upset early (66.7%).

Two of the eight upset victims were overseeded, based on KenPom rank; if you don't count overseeded teams, we have 10 teams and 4 of them (40.0%) won four or more games.

Historically, 42.2% of 1-seeds won four or more games, and 57.8% were upset early.


Overall, it looks like "slow" top 20 KenPom teams that are not overseeded and are a #1 or #2 appear to perform about the same as the historical average for those seeds. "Slow" top 20 KenPom teams that are seeded #3 or worse appear to perform much better than the historical average for those seeds.

Sample size and all that, but based on this data, it seems probable that while these slow teams may "appear over-rated" by KenPom, they probably aren't. So either he has "fixed" the "Wisconsin problem," or there never was one to begin with.

[uh_no]

Overall, it looks like "slow" top 20 KenPom teams that are not overseeded and are a #1 or #2 appear to perform about the same as the historical average for those seeds. "Slow" top 20 KenPom teams that are seeded #3 or worse appear to perform much better than the historical average for those seeds.

Sample size and all that, but based on this data, it seems probable that while these slow teams may "appear over-rated" by KenPom, they probably aren't. So either he has "fixed" the "Wisconsin problem," or there never was one to begin with.

were you using the original rankings or currently listed ones after he redid his whole formula a few years ago?

[Wahoo2000]

Overall, it looks like "slow" top 20 KenPom teams that are not overseeded and are a #1 or #2 appear to perform about the same as the historical average for those seeds. "Slow" top 20 KenPom teams that are seeded #3 or worse appear to perform much better than the historical average for those seeds.

Sample size and all that, but based on this data, it seems probable that while these slow teams may "appear over-rated" by KenPom, they probably aren't. So either he has "fixed" the "Wisconsin problem," or there never was one to begin with.

I can't lie, the Virginia fan in me was a little aroused by the bolded portion. I've been semi-ashamed of our postseason performance the last 4 seasons, and this at least provides some hope that our style doesn't make the level of success I'm looking for impossible.

[Kedsy]

were you using the original rankings or currently listed ones after he redid his whole formula a few years ago?

The currently listed ones. The observation I responded to was about this year's rankings, so I didn't want to get into a debate of "well it was broken before but now it's fixed," and thus I used the most recent material. (Also, as I said, I used his posted post-tourney rankings, rather than the (original) pre-tourney rankings.)

[TheOldBattleship]

(see above couple of posts)

It's not a bad idea, but it's not trivial to do in a clean way.

Ben Falk (formerly of the Sixers and Blazers front offices) filters garbage time out of the stats on his new site (https://cleaningtheglass.com/; also filters out end of quarter/game heaves). Here's his definition, which is, I think, useful to look at even though there's clearly a difference here between the NBA and college idea of "garbage time":

"GARBAGE TIME
Cleaning the Glass uses a definition of garbage time that is as objective as possible and generally matches up with most people's perception of when garbage time starts: when the game is out of hand, both teams have subbed out most of their starters, and the game never gets close again.

THE GRITTY DETAILS
The exact definition CTG uses is: the game has to be in the 4th quarter, the score differential has to be >= 25 for minutes 12-9, >= 20 for minutes 9-6, and >= 10 for the remainder of the quarter. Additionally, there have to be two or fewer starters on the floor combined between the two teams.

Importantly, the game can never go back to being non-garbage time, or this clock resets. For example, if it's a 30 point game to start the 4th quarter, but one team comes back and pulls the game within 8, that comeback is not counted as garbage time. If the leading team regains control and expands the lead back out, garbage time would start when the score went back above 10.

This might not capture all of what we'd call garbage time, but it seems important to err on the side of caution and not mistakenly filter out any game time that we would not consider garbage time."

[uh_no]

Ben Falk (formerly of the Sixers and Blazers front offices) filters garbage time out of the stats on his new site (https://cleaningtheglass.com/; also filters out end of quarter/game heaves). Here's his definition, which is, I think, useful to look at even though there's clearly a difference here between the NBA and college idea of "garbage time":

"GARBAGE TIME
Cleaning the Glass uses a definition of garbage time that is as objective as possible and generally matches up with most people's perception of when garbage time starts: when the game is out of hand, both teams have subbed out most of their starters, and the game never gets close again.

THE GRITTY DETAILS
The exact definition CTG uses is: the game has to be in the 4th quarter, the score differential has to be >= 25 for minutes 12-9, >= 20 for minutes 9-6, and >= 10 for the remainder of the quarter. Additionally, there have to be two or fewer starters on the floor combined between the two teams.

Importantly, the game can never go back to being non-garbage time, or this clock resets. For example, if it's a 30 point game to start the 4th quarter, but one team comes back and pulls the game within 8, that comeback is not counted as garbage time. If the leading team regains control and expands the lead back out, garbage time would start when the score went back above 10.

This might not capture all of what we'd call garbage time, but it seems important to err on the side of caution and not mistakenly filter out any game time that we would not consider garbage time."

decently fair definition.

[Bluedog]

Okay, I know that there are much better statistical measures of a season that RPI.

But the RPI remains the selection committee's favorite tool, so we need to pay attention to it.

I also know that it's too early to pay much attention to RPI rankings (or anything else, such as Ken Pom). Still, just for fun, I checked out the RPI as of today:

https://www.teamrankings.com/ncb/rpi/

Duke is 19 today ... and it's not likely to get a lot better until ACC play starts,

Duke's upcoming opponents are not all that -- Portland State is 339, Butler is 106, Texas is 98, Florida is No. 7 in the nation in the AP and No. 248 in RPI. Gonzaga is 148. Indiana is 203.

I take those rankings of evidence as to how the early season RPI can be warped. Still, it will be interesting to watch the rankings evolve as we get deeper into the season.


I expect our strength of schedule at the end of the season to be pretty bad, and in turn our RPI. Weak OOC schedule outside of MSU, and a favorable conference schedule. Which is good in the sense that we’ll take fewer losses, but bad in terms of computer numbers.


RPI Forecast uses Sagarin to project the season, and Duke ends up as #3 RPI and #17 SOS (http://www.rpiforecast.com/index2.html). Not bad at all. We have enough tough games on the schedule. As mentioned elsewhere, I don't think the ACC will suffer through a down year.

Duke is currently #1 in RPI. ;) (Not that it really means much, but just find it interesting in light of the above comments).
https://www.teamrankings.com/ncb/rpi/

[DevilFalcon]

After last night our offense is up even more, to where we are 2.6 above Villanova in second place. Our defense being #50 isn't going to cut it though. Need to get that in the 20s.

[uh_no]

Duke is currently #1 in RPI. ;) (Not that it really means much, but just find it interesting in light of the above comments).
https://www.teamrankings.com/ncb/rpi/

yeah. duke should be top in almost any system that doesn't consider margin of victory

[ChillinDuke]

Interesting tidbits to me...

KP lists Grayson Allen with a sensational oRTG of 139.7, good for 17th in the nation among all players, not just qualifiers. What is of further interest to me is that KP still lists G as a "Role Player" at 16-20% possession usage. For reference, Trae Young is #1 in the country in oRTG at 130.3 for all players with at least 28% usage. Not coincidentally, KP lists Trae as his early frontrunner for NPOY.

<snip>

- Chillin

Following up on this post I made a few days ago, Grayson had a hyper-efficient game last night against SFU. His oRTG is now 141.9, good for 12th in the nation among all players. I get his "usage" isn't as high as others. But he's still hyper efficient so far.

I'm torn if this post belongs in Dork Polls or Grayson Allen Points Chase (which is his de facto tribute thread). Probably Dork Polls though.

- Chillin

[Skitzle]

I hate seeing Duke at 54 on defense in kenpom. Then I remember that they were 59th with a def eff of 96.1 going into the tourney and ended up at 11th with 92 def eff in 2015

Only takes a few weeks... That's what I keep telling myself

Current 54 with def eff of 97...

[uh_no]

I hate seeing Duke at 54 on defense in kenpom. Then I remember that they were 59th with a def eff of 96.1 going into the tourney and ended up at 11th with 92 def eff in 2015

Only takes a few weeks... That's what I keep telling myself

Current 54 with def eff of 97...

bit of a survivor bias there. we have plenty of examples of the defense NOT turning it around.

That said, it seems this team CAN play defense at times...but is not yet consistent....that makes me more hopeful than anything else

[JasonEvans]

That said, it seems this team CAN play defense at times...but is not yet consistent...that makes me more hopeful than anything else

As I said on the podcast this past week, teaching a team that does something right part of the time to do it right more of the time is a heck of a lot easier than teaching a team that does not know how to do something, how to do it.

Does that make sense?

[Skitzle]

I hate seeing Duke at 54 on defense in kenpom. Then I remember that they were 59th with a def eff of 96.1 going into the tourney and ended up at 11th with 92 def eff in 2015

Only takes a few weeks... That's what I keep telling myself

Current 54 with def eff of 97...


bit of a survivor bias there. we have plenty of examples of the defense NOT turning it around.


I repeat "Only takes a few weeks... That's what I keep telling myself"

[Olympic Fan]

I've never been a big fan of ESPN's in-house polls -- the BPI in football and their Basketball Power Rating. Too often, their ratings don't make much sense.

Yesterday's new Basketball Ratings gave me pause:

http://www.espn.com/mens-college-basketball/story/_/id/21861294/college-basketball-power-rankings-curious-case-kentucky-wildcats

North Carolina was No. 5 the week before. Last week, they lost AT HOME to No. 196 (at the time in KenPom) Wofford, then beat No. 45 Ohio State on a neutral court.

This week, UNC dropped to No. 8 in the ESPN ratings.

Miami was No. 6 in the ratings. The 'Canes beat No. 213 Hawaii ON THE ROAD, the lost to No. 80 New Mexico State on a neutral court (Monday night's win didn't factor in for the 'Canes).

This week, Miami dropped to No. 22 in the ESPN ratings.

I can't figure out why Miami's week was so much worse than UNC's week. Certainly UNC over Ohio State was a better win than Miami over Hawaii (even on the road), but UNC's loss to Wofford was worse -- much worse -- than Miami's loss to NM State.

PS The comments in the article -- mentioning how weak Miami's schedule has been -- might be valid -- BUT you only just noticed? It was the same schedule last week when the 'Canes were No. 6 in your power ratings.

How does UNC drop 3 places ... while Miami drops 16 places?

[hudlow]

How does UNC drop 3 places ... while Miami drops 16 places?


They've had a hard year. They're feeling sensitive and fragile?

[uh_no]

with UNC's poor play of late, we are now projected to win every remaining game on our schedule.

[Billy Dat]

Teams that can manage to be KenPom Top 15/Top 20 in both offense and defense are usually well positioned to make a serious run at the brass ring. At this point, there are very few who qualify, elite on one end is tending to be paired with non-elite on the other end. There are some current exceptions:

Purdue - 5th offense, 9th defense
Kansas - 8th offense, 19th defense
MSU - 12th offense, 2nd defense

Then you've got Wichita State, West Virginia and UNC lurking near top 20 for each.

[Kedsy]

Teams that can manage to be KenPom Top 15/Top 20 in both offense and defense are usually well positioned to make a serious run at the brass ring. At this point, there are very few who qualify, elite on one end is tending to be paired with non-elite on the other end. There are some current exceptions:

Purdue - 5th offense, 9th defense
Kansas - 8th offense, 19th defense
MSU - 12th offense, 2nd defense

Then you've got Wichita State, West Virginia and UNC lurking near top 20 for each.

No offense, but this is basically a myth. Since we're talking about "well positioned," it follows we have to use pre-tournament data. I have 9 seasons worth of pre-T data, and in those nine seasons, four of the nine champions (44%) were in Pomeroy's top 20 in both offense and defense going into the Tournament. I suppose one could argue that almost half can be considered "well positioned" (though personally I'd disagree) so let's look further:

In those same nine seasons, there were 52 teams that were in Pomeroy's top 20 in both offense and defense going into the NCAAT. Of those:

5 (9.6%) got knocked out in the round of 64;
12 (23.1%) got knocked out in the round of 32;
11 (21.2%) got knocked out in the round of 16;
13 (25.0%) got knocked out in the round of 8;
10 (19.2%) made the Final Four (of those, 4 lost their Final Four game; 2 lost in the championship; and 4 won the championship).

Depending on how you define "lurking," it gets worse. For example, of the three teams that were in the top 21 in both categories (but not in top 20 for both, in other words top 20 on one side and 21st on the other), none of the three teams made it past the second round.

For comparison's sake, in those same nine seasons, there were 72 teams with #1 or #2 seeds. Of those, seven won championships (78%), and:

4 (5.6%) got knocked out in the round of 64;
15 (20.1%) got knocked out in the round of 32;
13 (18.1%) got knocked out in the round of 16;
21 (29.2%) got knocked out in the round of 8;
19 (26.4%) made the Final Four (of those, 7 lost their Final Four game; 5 lost in the championship; and 7 won the championship).

Of course, a lot of those #1 and #2 seeds were also ranked in Pomeroy's top 20 of both offense and defense, so there is a fair amount of overlap here. But I conclude that being in the top 20 for both doesn't give you any advantage over getting a top 2 seed and not being ranked in the top 20 for both.

[uh_no]

No offense, but this is basically a myth. Since we're talking about "well positioned," it follows we have to use pre-tournament data. I have 9 seasons worth of pre-T data, and in those nine seasons, four of the nine champions (44%) were in Pomeroy's top 20 in both offense and defense going into the Tournament. I suppose one could argue that almost half can be considered "well positioned" (though personally I'd disagree) so let's look further:

In those same nine seasons, there were 52 teams that were in Pomeroy's top 20 in both offense and defense going into the NCAAT. Of those:

5 (9.6%) got knocked out in the round of 64;
12 (23.1%) got knocked out in the round of 32;
11 (21.2%) got knocked out in the round of 16;
13 (25.0%) got knocked out in the round of 8;
10 (19.2%) made the Final Four (of those, 4 lost their Final Four game; 2 lost in the championship; and 4 won the championship).

Depending on how you define "lurking," it gets worse. For example, of the three teams that were in the top 21 in both categories (but not in top 20 for both, in other words top 20 on one side and 21st on the other), none of the three teams made it past the second round.

For comparison's sake, in those same nine seasons, there were 72 teams with #1 or #2 seeds. Of those, seven won championships (78%), and:

4 (5.6%) got knocked out in the round of 64;
15 (20.1%) got knocked out in the round of 32;
13 (18.1%) got knocked out in the round of 16;
21 (29.2%) got knocked out in the round of 8;
19 (26.4%) made the Final Four (of those, 7 lost their Final Four game; 5 lost in the championship; and 7 won the championship).

Of course, a lot of those #1 and #2 seeds were also ranked in Pomeroy's top 20 of both offense and defense, so there is a fair amount of overlap here. But I conclude that being in the top 20 for both doesn't give you any advantage over getting a top 2 seed and not being ranked in the top 20 for both.

yeah, i think the only reason you'd want to be good at both is because you're better being good at something than not being good at it. I'm skeptical that there is any sort of "bonus" to being marginally good at both over being excellent at one and marginal at the other.

If one believed there was a bonus, it would invalidate much of dork-poll theory...namely that your quality as a team can be derived from the independent offense and defense components. KP moved in that direction further last year when he moved from pythagorean based rankings to linear ones.

[Olympic Fan]

Got a chance to see the official RPI report today (which may be slightly different than ESPN's version.

Some real interesting stuff:

Duke is No. 1 ... Virginia is No. 2 ... Clemson is No. 5 ... UNC is No. 8. Then it drops to Miami at No. 22, FSU at No. 27, Syracuse No. 31, Louisville, No. 32 Notre Dame No. 37.

The bottom ACC teams have all made significant jumps recently -- Georgia Tech shot up from No. 245 to No. 197 last week; NC State from No. 145 to No. 99.

Obviously early and things will change in the next two months -- but not as much as you'd think. ACC teams are basically ranked so high that playing ACC opponents will keep the ACC teams ranked very high.

And I keep saying -- even though many of us prefer other rankings (KenPom, BPI, Sagarin, etc.), RPI has always been the selection committee's favorite tool.

[DallasDevil]

Got a chance to see the official RPI report today (which may be slightly different than ESPN's version.

Some real interesting stuff:

Duke is No. 1 ... Virginia is No. 2 ... Clemson is No. 5 ... UNC is No. 8. Then it drops to Miami at No. 22, FSU at No. 27, Syracuse No. 31, Louisville, No. 32 Notre Dame No. 37.

The bottom ACC teams have all made significant jumps recently -- Georgia Tech shot up from No. 245 to No. 197 last week; NC State from No. 145 to No. 99.

Obviously early and things will change in the next two months -- but not as much as you'd think. ACC teams are basically ranked so high that playing ACC opponents will keep the ACC teams ranked very high.

And I keep saying -- even though many of us prefer other rankings (KenPom, BPI, Sagarin, etc.), RPI has always been the selection committee's favorite tool.

It will be interesting to see how the changes in the committee's team sheets for this year impact how they view games. For example, Texas is currently 59 and Boston College is currently 76. If NC State and BC could get up to the top 75 by the end of the year, those losses would be quadrant 1 losses and would not look as bad to the committee. Similarly, if Texas can break into the top 50 then that would be another quadrant 1 win. Two current quadrant 1 wins (Florida, 48, and FSU, 27) are at risk of becoming less valuable quadrant 2 wins if those teams slip past 50 and 30, respectively. We play Syracuse, Louisville, and Notre Dame at home this year, so it would help if they all moved into the top 30.

[MChambers]

Duke's offense has slipped to #2 in Pomeroy, 0.1 behind Villanova.

It's over.

[uh_no]

Duke's offense has slipped to #2 in Pomeroy, 0.1 behind Villanova.

It's over.

it had slipped to #2 last week and then bounced back.

but yeah, a gazillion turnovers will do that.

[Troublemaker]

Interesting. After the Miami game, Duke's defense was ranked #72. Somehow, without playing a game between then and this ongoing Pitt game (it's halftime), Duke's defense is now ranked #85.

[Mtn.Devil.91.92.01.10.15]

Interesting. After the Miami game, Duke's defense was ranked #72. Somehow, without playing a game between then and this ongoing Pitt game (it's halftime), Duke's defense is now ranked #85.

Well, other games were played. Maybe we moved down by standing still?

[MChambers]

Interesting. After the Miami game, Duke's defense was ranked #72. Somehow, without playing a game between then and this ongoing Pitt game (it's halftime), Duke's defense is now ranked #85.

MSU destroyed Indiana, so that probably contributed.

[duke4ever19]

MSU destroyed Indiana, so that probably contributed.

Does KenPom take into account injuries? Mich St. was also at home vs Indiana and down a starter in Deron Davis (out for the year), who scored 16 against us.

I'd genuinely like to know how KenPom takes stuff like injuries this into account (if he does), because that certainly affects a team's chemistry and often how many games they might win minus a starting player.

Edit: We were briefly in the 80's on D today, but after today's win over Pitt, KenPom shows us back in the 70s on defense (78). Still a poorer rating than post-Miami, though.

[robed deity]

Being an obsessive kenpom checker, I noticed that the drop in D ranking (without playing) was due to the relatively poor recent offensive performance of Indiana, Wake Forest, and today, Texas.

[MChambers]

Does KenPom take into account injuries? Mich St. was also at home vs Indiana and down a starter in Deron Davis (out for the year), who scored 16 against us.

I'd genuinely like to know how KenPom takes stuff like injuries this into account (if he does), because that certainly affects a team's chemistry and often how many games they might win minus a starting player.

Edit: We were briefly in the 80's on D today, but after today's win over Pitt, KenPom shows us back in the 70s on defense (78). Still a poorer rating than post-Miami, though.

I don’t think KenPom takes injuries into account, at all. Not at all sure how any statistical system could.

Down to 79, as of 8:43 pm. Dripp8ng quickly.

[uh_no]

I don’t think KenPom takes injuries into account, at all. Not at all sure how any statistical system could.

Down to 79, as of 8:43 pm. Dripp8ng quickly.

i believe he does have some method for adjusting when a high usage player misses games....making them weigh less...but I can't seem to find a refernece.

[juise]

#3 in KenPom, but #1 in your hearts... the Duke Bluuuuue Devilllls!

It’s striking that Duke is currently exactly one behind Villanova in offensive, defensive, and overall efficency. I would not call the two teams terribly similar.

[ChillinDuke]

I'm not sure if this belongs here or deserves its own thread (or is being talked about elsewhere), but quietly (and somewhat remarkably, at least to me) KenPom has a new Dukie in his kPOY Top-10.

And it's not Grayson Allen.

Wendell Carter is now ranked #10, right behind Villanova's Mikal Bridges. Wow!

Marvin Bagley is still ranked #3, behind #2 Trae Young and #1 Jock Landale.

For me, this really puts an exclamation point on the often-discussed topic "If Marvin Bagley wasn't on Duke, everyone would be talking about Wendell Carter as one of the best players in the country."

- Chillin

[CDu]

I'm not sure if this belongs here or deserves its own thread (or is being talked about elsewhere), but quietly (and somewhat remarkably, at least to me) KenPom has a new Dukie in his kPOY Top-10.

And it's not Grayson Allen.

Wendell Carter is now ranked #10, right behind Villanova's Mikal Bridges. Wow!

Marvin Bagley is still ranked #3, behind #2 Trae Young and #1 Jock Landale.

For me, this really puts an exclamation point on the often-discussed topic "If Marvin Bagley wasn't on Duke, everyone would be talking about Wendell Carter as one of the best players in the country."

- Chillin

Yep. It's really hard to have two players dominate inside consistently. Not that both can't play well together, but it's just hard to see the statistical results for both. Largely, they are competing for a lot of the same stats. Teams just aren't going to top 50% of offensive rebound opportunities consistently, nor 80% of defensive rebounds, so those rebound numbers are to some degree coming at the expense of our other star inside. And Bagley is the slightly superior talent, so he's been the one getting the pub. But Carter is REALLY freaking good too. And I am glad he's getting to show that off right now.

[Listen to Quants]

I don’t think KenPom takes injuries into account, at all. Not at all sure how any statistical system could.

Down to 79, as of 8:43 pm. Dripp8ng quickly.

Late to this, but if you are still interested, the ESPN BPI index does some accounting for injuries. What they do is discount the games that have important players out.


Dean Oliver wrote: ".... If a team or its opponent is missing one of its most important players (determined by minutes per game) for a contest, that game is less important for ranking the teams compared to games in which both teams are at full strength. ...."

[tbyers11]

Late to this, but if you are still interested, the ESPN BPI index does some accounting for injuries. What they do is discount the games that have important players out.


Dean Oliver wrote: "... If a team or its opponent is missing one of its most important players (determined by minutes per game) for a contest, that game is less important for ranking the teams compared to games in which both teams are at full strength. ..."

So by this metric BPI would discount our recent upsurge a bit compared to a method that doesn't account for injury at all? Or does BPI only account for injury if the team is negatively impacted (to the whole)?

[Listen to Quants]

So by this metric BPI would discount our recent upsurge a bit compared to a method that doesn't account for injury at all? Or does BPI only account for injury if the team is negatively impacted (to the whole)?

I was digging out an old paragraph from Oliver (who is a quality sports-stats analyst), so it isn't clear to me what they do now. Still, it appears as you say, the only adjustment is the discount and BPI will discount the games without MBIII.

Obviously (at least I think), a better method would be to treat an injury as a total team adjustment, something like a home-neutral-away adjustment. That, however, would be very hard given the limited games played since you'd need a stable plus-minus stat or something like it.

[JasonEvans]

Reviving a dormant thread because this is the best place to put this info---

It is sometimes said that the national champion tends to almost always be a team that is top 10 (or really close to the top 10) in KenPom's offense and defense.

For example:


2017 - UNC 9 O, 11 D
2016 - Nova 3 O, 5 D
2015 - Duke 3 O, 11 D
2014 - UConn 39 O, 10 D (huge upset winner)
2013 - Louville 7 O, 1 D
2012 - Kentucky 2 O, 7 D
2011 - UConn 19 O, 15 D (as a 3 seed, they were again a bit of an upset winner)
2010 - Duke 1 O, 5 D
2009 - UNC 1 O, 1 D
2008 - Kansas 1 O, 2 D
2007 - Fla 1 O, 13 D
2006 - Fla 3 O, 7 D
2005 - UNC 2 O, 5 D
2004 - UConn 9 O, 6 D
2003 - Syrac 17 O, 14 D
2002 - Maryland 4 O, 7 D

So, if you count teams that were just barely (by 1-3 spots) outside the top 10, we have 13 out of 16 champs coming from teams who were top 10 (or really, really close to it) in both offense and defense.

At this point, right now, there are only 2 teams who qualify. Duke (3 O, 10 D) and Michigan St (11 O, 9 D). Now, I am well aware that the numbers change as teams march through the NCAA tourney and the team that wins the tourney almost always ends up improving their numbers over the course of the tournament, because they are winning games against top competition. But, at this moment, Pomeroy's rankings would say that Duke and Michigan St are the two teams who's profiles most closely match that of a national champion.

-Jason "me likey!" Evans

[proelitedota]

Reviving a dormant thread because this is the best place to put this info---

It is sometimes said that the national champion tends to almost always be a team that is top 10 (or really close to the top 10) in KenPom's offense and defense.

For example:


2017 - UNC 9 O, 11 D
2016 - Nova 3 O, 5 D
2015 - Duke 3 O, 11 D
2014 - UConn 39 O, 10 D (huge upset winner)
2013 - Louville 7 O, 1 D
2012 - Kentucky 2 O, 7 D
2011 - UConn 19 O, 15 D (as a 3 seed, they were again a bit of an upset winner)
2010 - Duke 1 O, 5 D
2009 - UNC 1 O, 1 D
2008 - Kansas 1 O, 2 D
2007 - Fla 1 O, 13 D
2006 - Fla 3 O, 7 D
2005 - UNC 2 O, 5 D
2004 - UConn 9 O, 6 D
2003 - Syrac 17 O, 14 D
2002 - Maryland 4 O, 7 D

So, if you count teams that were just barely (by 1-3 spots) outside the top 10, we have 13 out of 16 champs coming from teams who were top 10 (or really, really close to it) in both offense and defense.

At this point, right now, there are only 2 teams who qualify. Duke (3 O, 10 D) and Michigan St (11 O, 9 D). Now, I am well aware that the numbers change as teams march through the NCAA tourney and the team that wins the tourney almost always ends up improving their numbers over the course of the tournament, because they are winning games against top competition. But, at this moment, Pomeroy's rankings would say that Duke and Michigan St are the two teams who's profiles most closely match that of a national champion.

-Jason "me likey!" Evans

Reddit has a thread every week tracking what teams where at the current point of season and how that matches up to past championship teams at the same point in the season. AJO, AJD, Offensive rebounding are the biggest indicators to post-season success. I'll post the link the thread when it appears again.

[Kedsy]

Reviving a dormant thread because this is the best place to put this info---

It is sometimes said that the national champion tends to almost always be a team that is top 10 (or really close to the top 10) in KenPom's offense and defense.

For example:


2017 - UNC 9 O, 11 D
2016 - Nova 3 O, 5 D
2015 - Duke 3 O, 11 D
2014 - UConn 39 O, 10 D (huge upset winner)
2013 - Louville 7 O, 1 D
2012 - Kentucky 2 O, 7 D
2011 - UConn 19 O, 15 D (as a 3 seed, they were again a bit of an upset winner)
2010 - Duke 1 O, 5 D
2009 - UNC 1 O, 1 D
2008 - Kansas 1 O, 2 D
2007 - Fla 1 O, 13 D
2006 - Fla 3 O, 7 D
2005 - UNC 2 O, 5 D
2004 - UConn 9 O, 6 D
2003 - Syrac 17 O, 14 D
2002 - Maryland 4 O, 7 D

So, if you count teams that were just barely (by 1-3 spots) outside the top 10, we have 13 out of 16 champs coming from teams who were top 10 (or really, really close to it) in both offense and defense.

At this point, right now, there are only 2 teams who qualify. Duke (3 O, 10 D) and Michigan St (11 O, 9 D). Now, I am well aware that the numbers change as teams march through the NCAA tourney and the team that wins the tourney almost always ends up improving their numbers over the course of the tournament, because they are winning games against top competition. But, at this moment, Pomeroy's rankings would say that Duke and Michigan St are the two teams who's profiles most closely match that of a national champion.

-Jason "me likey!" Evans

If you're using the above to say that Duke has the potential to play like a champion, I guess that's ok. But otherwise, please stop.

I only have pre-tourney numbers since 2009, but they're nowhere close to the post-tourney numbers you cite. In fact, only 2 of the past 9 champs have been top 10 in both going into the NCAAT, and only one other was "really, really close":

2009 UNC: 1/35
2010 Duke 1/4
2011 UConn 21/31
2012 UK 2/6
2013 L'Ville 15/1
2014 UConn 80/11
2015 Duke 3/57
2016 V'Nova 11/7
2017 UNC 4/25

So it's not that teams improve a little in the tournament. The winners improve so much this stat is meaningless.

[Kedsy]

If you're using the above to say that Duke has the potential to play like a champion, I guess that's ok. But otherwise, please stop.

I only have pre-tourney numbers since 2009, but they're nowhere close to the post-tourney numbers you cite. In fact, only 2 of the past 9 champs have been top 10 in both going into the NCAAT, and only one other was "really, really close":

2009 UNC: 1/35
2010 Duke 1/4
2011 UConn 21/31
2012 UK 2/6
2013 L'Ville 15/1
2014 UConn 80/11
2015 Duke 3/57
2016 V'Nova 11/7
2017 UNC 4/25

So it's not that teams improve a little in the tournament. The winners improve so much this stat is meaningless.

Let me put this another way. In the last 9 years there have been 23 teams that met your criteria going into the tournament (top13 in both O & D) -- 15 1-seeds; 4 2-seeds; 1 3-seed; and 3 4-seeds.

Of those 23 teams, 3 won the championship; 2 lost in the final game; 2 lost in the final four; 6 lost in the elite 8; 5 lost in the sweet 16; and 5 lost in the round of 32. Basically this population did no better, probably a little worse, than you'd expect from a group of teams with that seed distribution.

In other words, the stat is bogus.

[proelitedota]

Reddit has a thread every week tracking what teams where at the current point of season and how that matches up to past championship teams at the same point in the season. AJO, AJD, Offensive rebounding are the biggest indicators to post-season success. I'll post the link the thread when it appears again.

Found the latest threads from a week ago. Jason this might make for good podcast material.

https://www.reddit.com/r/CollegeBasketball/comments/7z1x80/statistics_of_kenpom_era_champions_pre_tourney_vs/
https://www.reddit.com/r/CollegeBasketball/comments/7n37jy/kenpom_data_since_2002/

Probably the best one as it goes into detail over more categories than just AJO and AJD:
https://www.reddit.com/r/CollegeBasketball/comments/7y17fm/since_2008_3440_final_four_teams_ranked_in_the/

[curtis325]

Let me put this another way. In the last 9 years there have been 23 teams that met your criteria going into the tournament (top13 in both O & D) -- 15 1-seeds; 4 2-seeds; 1 3-seed; and 3 4-seeds.

Of those 23 teams, 3 won the championship; 2 lost in the final game; 2 lost in the final four; 6 lost in the elite 8; 5 lost in the sweet 16; and 5 lost in the round of 32. Basically this population did no better, probably a little worse, than you'd expect from a group of teams with that seed distribution.

In other words, the stat is bogus.

Nice analysis. It looks to me that Duke has a 30% chance of making the final 4. Small sample size notwithstanding, it is obviously a lock!

[Kedsy]

Nice analysis. It looks to me that Duke has a 30% chance of making the final 4. Small sample size notwithstanding, it is obviously a lock!

Even more of a lock if Duke gets a #1 seed because, historically, 42% of #1 seeds make the Final Four.

In the past 9 seasons, 12 #1-seeds have made the Final Four (33%). BUT when those #1-seeds were top 13 in both offense and defense, only 26.7% made the Final Four. For #1-seeds who were NOT top 13 in both offense and defense, 38.1% made the Final Four.

Here's the detail for those 9 seasons:

#1, top 13 in both O and D (15 teams):
2 champs (13.3%); 0 lost in final game (0%); 2 lost in Final Four (13.3%); 4 lost in Elite Eight (26.7%); 3 lost in Sweet 16 (20%); 4 lost in round of 32 (26.7%).

#1, NOT top 13 in both O and D (21 teams):
4 champs (19.0%); 2 lost in final game (9.5%); 2 lost in Final Four (9.5%); 7 lost in Elite Eight (33.0%); 4 lost in Sweet 16 (19.0%); 2 lost in round of 32 (9.5%)


Small sample sizes, but clearly once you're a #1-seed, also being top 13 in both O and D doesn't give you any kind of edge. Like I said, bogus stat.

[bedeviled]

For some simple fun with statistics:
Algebracket (http://algebracket.com/) allows you to predict bracket results by weighting the statistics you think are important

For advanced statisticians and modelers:
The annual Kaggle March Madness bracket challenge can be found here (https://www.kaggle.com/c/mens-machine-learning-competition-2018). Even if you don't participate, it is sometimes interesting to follow the discussions and outcomes (and hope that at least a little makes sense to plebs like me).

And, if you want to play around with statistics and modeling at home, Kaggle offers the following data for download (via an API, using Python):
Weekly team rankings from "dozens of top rating systems - Pomeroy, Sagarin, RPI, ESPN, etc" since 2002-03
Day, Score, and Home/Away/Neutral of all regular season and post-season games since 1984-85
City location for all regular season and post-season games since 2009-10
Box Scores for all regular season and post-season games since 2002-03
Play-by-play event logs (including plays by individual players) for 99% of regular season and post-season games since 2009-10
Seeds and Regions for NCAAT games since 1984-85
Day, Round, and Score of NCAAT games since 1984-85