#16 in the AP

[punaj]

http://collegebasketball.ap.org/poll/2017/17

Wisc the only other B1G team in the top 25... a few others receiving votes. I think this year's NCAAs are going to be fun to watch... I could see the B1G surprising people given the national press they have gotten this year.

 

[Dryfly88]

I have to catch some more Florida games before I fill out my tournament bracket. All of the metrics love them and so do the polls. KenPom and BPI both have them #7 and RPI is #6. I have seen them twice and thought they looked like a good team but no where close to the top 10. I normally really like the KenPom numbers this late in the season so maybe I've really missed something with the times I saw them??

 

[*4purdue*]

I have to catch some more Florida games before I fill out my tournament bracket. All of the metrics love them and so do the polls. KenPom and BPI both have them #7 and RPI is #6. I have seen them twice and thought they looked like a good team but no where close to the top 10. I normally really like the KenPom numbers this late in the season so maybe I've really missed something with the times I saw them??

I watched them for the first time against UK on Saturday. They looked okay. I don't think they looked great. They lost a center for the year apparently. If they're top 10, then so is Purdue IMO.

 

[TopSecretBoiler]

I have to catch some more Florida games before I fill out my tournament bracket. All of the metrics love them and so do the polls. KenPom and BPI both have them #7 and RPI is #6. I have seen them twice and thought they looked like a good team but no where close to the top 10. I normally really like the KenPom numbers this late in the season so maybe I've really missed something with the times I saw them??

you didn't miss anything. They haven't beaten anyone of note other than uk at home. Lousy record against the top 25. Not sure why kenpom is loving them so much.

 

[TopSecretBoiler]

the zags getting 1st place votes is hysterical. I'm making voodoo dolls to get Purdue into their 1 seed bracket.

 

[arcb102000]

I have also watched Florida a few times. I do not get the Gator love. I look for an early exit for them.

 

[FirstDownB]

I have to catch some more Florida games before I fill out my tournament bracket. All of the metrics love them and so do the polls. KenPom and BPI both have them #7 and RPI is #6. I have seen them twice and thought they looked like a good team but no where close to the top 10. I normally really like the KenPom numbers this late in the season so maybe I've really missed something with the times I saw them??

Seems to be a combination of SOS and scoring margin.

Florida SOS is artificially high due to avoiding playing the worst of the worst cupcakes. Any tourney quality team can easily blow out #240 team just as easily as the #340 team, but one of them drives your metric much farther down. Smart/lucky scheduling can manipulate your SOS without much effect on win probability.

My immediate tweak to RPI would be to count only games against top 150 teams as a "tournament RPI". Throw out all the cupcakes (unless you lose, perhaps).

 

[cprh9u]

Seems to be a combination of SOS and scoring margin.

Florida SOS is artificially high due to avoiding playing the worst of the worst cupcakes. Any tourney quality team can easily blow out #240 team just as easily as the #340 team, but one of them drives your metric much farther down. Smart/lucky scheduling can manipulate your SOS without much effect on win probability.

My immediate tweak to RPI would be to count only games against top 150 teams as a "tournament RPI". Throw out all the cupcakes (unless you lose, perhaps).

I think lucky is a big factor. It's not easy to decipher between the 240 and 340 team a few years in advance, especially not when there are several cupcakes that can give you a difficult matchup and potentially beat you if you are trying to get too cute.

 

[BoilerStation]

I thought somebody stated on here that Florida's arena was under construction or repair or something for the nonconf schedule and they played all their games away or at neutral sites instead of at home which boosted their numbers.

 

[Dryfly88]

I thought somebody stated on here that Florida's arena was under construction or repair or something for the nonconf schedule and they played all their games away or at neutral sites instead of at home which boosted their numbers.

Good point. I looked that up and their first 10 games were away/neutral. They opened the "new" arena on 12/21 this year. Thanks for bringing that up.

http://www.onlygators.com/category/basketball/

 

[blockm2]

Florida SOS is artificially high due to avoiding playing the worst of the worst cupcakes. Any tourney quality team can easily blow out #240 team just as easily as the #340 team, but one of them drives your metric much farther down. Smart/lucky scheduling can manipulate your SOS without much effect on win probability.

Sure, that explains some of RPI's quirks, but not KenPom. He even limits the effect of blowout wins to some degree. I think a lot of what goes into Florida's rankings is that they have had some very close losses against very good competition. Gonzaga by 5 on a neutral court, FSU by 5 on the road, South Carolina by 4 on the road, Kentucky by 10 on the road, etc. Did they lose to those teams? Sure, but close losses against really good opponents mean you are probably a really good team. Much more so than a win over a mediocre opponent. Florida combines a good average margin of victory with a strong SOS (only 3 games total vs sub 171 in KenPom).

 

[FirstDownB]

I think lucky is a big factor. It's not easy to decipher between the 240 and 340 team a few years in advance, especially not when there are several cupcakes that can give you a difficult matchup and potentially beat you if you are trying to get too cute.

To some degree..

But I submit these two lists:

A: Mercer (1- 2014), North Florida (1- 2015), Charlotte (0- 2005), Florida Gulf Coast (2- 2016), Little Rock (2- 2016)

B: Cleveland State (1- 2009), Norfolk State (1- 2012), NJIT (0- N/A), McNeese (0- 2002), Western Illinois (0- N/A)

List A is Florida's bottom 5 OOC opponents with # of NCAAT appearances in the last 10 years and last year of appearance. List B is Purdue's.
Based on nothing more than recent tourney appearances (didn't have time to compile winning %), one could have comfortably predicted that Florida wasn't going to face too many 240+ opponents this season.

Florida scheduled aggressively and will be rewarded for it.

 

[FirstDownB]

Sure, that explains some of RPI's quirks, but not KenPom. He even limits the effect of blowout wins to some degree. I think a lot of what goes into Florida's rankings is that they have had some very close losses against very good competition. Gonzaga by 5 on a neutral court, FSU by 5 on the road, South Carolina by 4 on the road, Kentucky by 10 on the road, etc. Did they lose to those teams? Sure, but close losses against really good opponents mean you are probably a really good team. Much more so than a win over a mediocre opponent. Florida combines a good average margin of victory with a strong SOS (only 3 games total vs sub 171 in KenPom).

You cut off my first line:
Seems to be a combination of SOS and scoring margin.

When you have a handful of close losses and a lot of blowout wins, by default you will tend to have a high scoring margin.

 

[blockm2]

You cut off my first line:
Seems to be a combination of SOS and scoring margin.

When you have a handful of close losses and a lot of blowout wins, by default you will tend to have a high scoring margin.

I know I did. You just seemed to imply Florida was essentially gaming the system of KenPom by not playing of patsies. They aren't. Unlike RPI, KenPom does a far better job of appropriately weighting games against horrible opponents.

RPI can be manipulated by scheduling bad, not terrible, competition since it doesn't care about scoring margin. KenPom is the opposite. It doesn't care about wins and losses and it basically only cares about scoring margin. Florida is fortunate that KenPom is far more important for predicting future wins and losses.

 

[JohnnyDoeBoiler]

I think lucky is a big factor. It's not easy to decipher between the 240 and 340 team a few years in advance, especially not when there are several cupcakes that can give you a difficult matchup and potentially beat you if you are trying to get too cute.

Sure it is....don't play teams from traditional horrible conferences. Play teams from the MAC, Atlantic Sun, Atlantic 10, Colonial Athletic, Horizon, Missouri Valley, and the WAC. Stay way from conferences where their winners are traditionally those that get the 15-16 seeds every single year. I am guessing if you did that, you would find yourself playing a lot of 100-175 games instead of the dreaded 200-350 games that do absolutely nothing for RPI. If Purdue had done that this year, we probably wouldn't have seen the number of blowouts BUT the RPI would be much higher than it currently is. I think that type of schedule would do Purdue much better than what we have seen over the last few years. Obviously this years was better with Nova, ND,Lousiville and the tourney giving us Auburn/Arizona State...but other than that the teams were pretty weak. I am not saying go off the deep end like Izzo did with MSU this season and suddenly schedule 8 non-conference games against the top 50 from the previous year...BUT getting more in to that 100-175 zone would be much better IMO.

 

[BoilerStation]

Didn't the NCAA say RPI wouldn't be playing nearly as big a factor going forward after this season and they were looking into other rating/ranking systems so to talk about what we should do in the future in relevance to the RPI may be moot. I'd just like to see games scheduled against the top 5 conferences each season. We know we will get an ACC team, possibly 2 depending on if we are playing Butler/ND and then throw in Big 12/SEC/Big East/Pac 12 teams. I liked playing WVU and Louisville in the past. They don't have to be top tier teams but just get one from each of the conferences then fill the rest in with some lesser teams to polish and experiment against.

 

[FirstDownB]

I know I did. You just seemed to imply Florida was essentially gaming the system of KenPom by not playing of patsies. They aren't. Unlike RPI, KenPom does a far better job of appropriately weighting games against horrible opponents.

RPI can be manipulated by scheduling bad, not terrible, competition since it doesn't care about scoring margin. KenPom is the opposite. It doesn't care about wins and losses and it basically only cares about scoring margin. Florida is fortunate that KenPom is far more important for predicting future wins and losses.

To be clear, they are gaming the RPI, which is the primary metric used by the committee.

But I believe there is also error in any SOS calculation that works its way into the very best of metrics such as KenPom. The algorithms are not sophisticated enough to distinguish, for example, between the following simplified examples:
1) Beating the #1 team by 10 points and the #340 team by 20 points (+15 avg scoring margin and 170.5 avg opponent rank)
2) Beating the #170 and #171 teams by 15 points apiece (+15 avg scoring margin and 170.5 avg opponent rank)

In reality it is much more difficult to do #1 in part because there is a far greater gap between teams #1 and #170 than teams #170 and #340. But the algorithm simply averages the numbers.

A game control metric is perhaps more accurate than either. A game where a team is up by 30 at half time, pulls its starters with 8 minutes left and wins by 20 is much different than a game where a team is up by 8 at half time, leaves in its starters the whole game and wins by 20. But not to KenPom.

 

[SIBoiler2]

Didn't the NCAA say RPI wouldn't be playing nearly as big a factor going forward after this season and they were looking into other rating/ranking systems so to talk about what we should do in the future in relevance to the RPI may be moot. I'd just like to see games scheduled against the top 5 conferences each season. We know we will get an ACC team, possibly 2 depending on if we are playing Butler/ND and then throw in Big 12/SEC/Big East/Pac 12 teams. I liked playing WVU and Louisville in the past. They don't have to be top tier teams but just get one from each of the conferences then fill the rest in with some lesser teams to polish and experiment against.

They're going to address that RPI thing right after the final report comes out on the UNC situation.

 

[TopSecretBoiler]

They're going to address that RPI thing right after the final report comes out on the UNC situation.

aw snap! so true...so true...

 

[FirstDownB]

Sure it is....don't play teams from traditional horrible conferences. Play teams from the MAC, Atlantic Sun, Atlantic 10, Colonial Athletic, Horizon, Missouri Valley, and the WAC. Stay way from conferences where their winners are traditionally those that get the 15-16 seeds every single year. I am guessing if you did that, you would find yourself playing a lot of 100-175 games instead of the dreaded 200-350 games that do absolutely nothing for RPI. If Purdue had done that this year, we probably wouldn't have seen the number of blowouts BUT the RPI would be much higher than it currently is. I think that type of schedule would do Purdue much better than what we have seen over the last few years. Obviously this years was better with Nova, ND,Lousiville and the tourney giving us Auburn/Arizona State...but other than that the teams were pretty weak. I am not saying go off the deep end like Izzo did with MSU this season and suddenly schedule 8 non-conference games against the top 50 from the previous year...BUT getting more in to that 100-175 zone would be much better IMO.

I can even live with most of those conferences. It's the MEACs, SWACs, and Southlands that kill your SOS. Avoid historically bottom feeder teams from bottom feeder conferences. IMO there are absolutely no metrics that fairly assess the outcomes of games where there is a tremendous mismatch.

Purdue bottom 4 opponents: (KenPom rank)
McNeese (317) W 109-65
NJIT (274) W 79-68
W. Illinois (298) W 82-50
Norfolk St. (295) W 91-45

Average KenPom rank= 296.75
Average scoring margin= +33.25

Next 4 from the bottom opponents: (KenPom rank)
Utah St (135) W 85-64
Morehead St. (199) W 90-56
Cleveland St. (228) W 77-53
Az St. (132) W 97-64

Average KenPom rank= 173.5
Average scoring margin= +28.0

Here it is demonstrated quite clearly that for the difference in KenPom SOS of 123 places between the 2 groups, we only beat the bottom 4 group by an average of 5 points more.

The only other solution would be to run up the score on bottom feeders by as much as possible, which is not only considered poor sportsmanship, but also unnecessarily risky.

 

[delish1]

They're going to address that RPI thing right after the final report comes out on the UNC situation.

I heard an interview with the committee chair somewhere that said the committee had as many questions as answers after that Summitt with all the analytic guys

I think rpi will still be around at least one more year. Purdue and the big ten do need to do a better job of gaming it like the big 12, mvc, and mountain west have done for years

 

[FirstDownB]

I heard an interview with the committee chair somewhere that said the committee had as many questions as answers after that Summitt with all the analytic guys

I think rpi will still be around at least one more year. Purdue and the big ten do need to do a better job of gaming it like the big 12, mvc, and mountain west have done for years

I believe one of the main sticking points was if and how to address scoring margin without changing the way games are played with regard to sportsmanship, risking unnecessary injury, etc.
I agree the NCAA is going to be slow to move on anything with scoring margin.. which is why in the meantime I would like to see an enhanced RPI, a "tournament RPI" used for selection and seeding purposes. That way you can just throw out those games against less competitive teams as if they never happened. This would be such a simple move it could be done next year.

 

[Purdue Grad in Texas]

I believe one of the main sticking points was if and how to address scoring margin without changing the way games are played with regard to sportsmanship, risking unnecessary injury, etc.
I agree the NCAA is going to be slow to move on anything with scoring margin.. which is why in the meantime I would like to see an enhanced RPI, a "tournament RPI" used for selection and seeding purposes. That way you can just throw out those games against less competitive teams as if they never happened. This would be such a simple move it could be done next year.

At a minimum, I've advocated throwing out the lowest......essentially giving one "free bite" to offset scheduling quirks, advanced timetables, etc. Hopefully, it doesn't come to needing a wooden stake or some exorcism to get rid of its lingering effects.

 

[ARSmitty30]

the zags getting 1st place votes is hysterical. I'm making voodoo dolls to get Purdue into their 1 seed bracket.

I just wanna thank you, from the bottom of my heart, for posting this graphic. I've never seen it before and I'm not sure why. Perhaps I'm just a bit slap happy right now, but the Jobu playing card is literally make my eyes water with laughter. Wow. Thank you.

 

[blockm2]

But I believe there is also error in any SOS calculation that works its way into the very best of metrics such as KenPom. The algorithms are not sophisticated enough to distinguish, for example, between the following simplified examples:
1) Beating the #1 team by 10 points and the #340 team by 20 points (+15 avg scoring margin and 170.5 avg opponent rank)
2) Beating the #170 and #171 teams by 15 points apiece (+15 avg scoring margin and 170.5 avg opponent rank)

In reality it is much more difficult to do #1 in part because there is a far greater gap between teams #1 and #170 than teams #170 and #340. But the algorithm simply averages the numbers.

I disagree quite strongly. KenPom doesn't "average the numbers". It compares your performance vs what was expected and then rates it per possession (depending on the number of possessions in the game). It does this for each game. It also doesn't call the difference between #1 and #170 as equal to the difference between #170 and #340.

For example, #1 Gonzaga currently rates @ + 32 (points per 100 possessions, round off to whole numbers). #170 Loyola Marymount is 0. #340 Alabama State is -20. So #170 is much, much closer to #340 than they are to #1. In fact, team #101 is about 1/2 way between #1 and #340 in terms of adjusted points per 100 possessions.


And why do I think this works so well for KenPom? Because it predicts with >95% accuracy the Vegas point spread. I think anybody would agree that the Vegas line is about the best ranking system you will get between teams because if it wasn't the casinos would be open to losing money on it to smart, well funded bettors. KenPom really isn't far back and for all intents and purposes can accurately approximate it if you want to compare teams to each other.

 

[FirstDownB]

I disagree quite strongly. KenPom doesn't "average the numbers". It compares your performance vs what was expected and then rates it per possession (depending on the number of possessions in the game). It does this for each game. It also doesn't call the difference between #1 and #170 as equal to the difference between #170 and #340.

For example, #1 Gonzaga currently rates @ + 32 (points per 100 possessions, round off to whole numbers). #170 Loyola Marymount is 0. #340 Alabama State is -20. So #170 is much, much closer to #340 than they are to #1. In fact, team #101 is about 1/2 way between #1 and #340 in terms of adjusted points per 100 possessions.


And why do I think this works so well for KenPom? Because it predicts with >95% accuracy the Vegas point spread. I think anybody would agree that the Vegas line is about the best ranking system you will get between teams because if it wasn't the casinos would be open to losing money on it to smart, well funded bettors. KenPom really isn't far back and for all intents and purposes can accurately approximate it if you want to compare teams to each other.

I think we're getting lost in technicalities of the algorithms. To what do you attribute Florida's high KenPom rating?

 

[blockm2]

I think we're getting lost in technicalities of the algorithms. To what do you attribute Florida's high KenPom rating?

they've played a good schedule and performed well. They currently rate #5 in RPI, #6 in BPI, #8 in Sagarin, #12 in AP poll, and #12 in Coaches poll.

The better question is why wouldn't they be also rated highly in KenPom?

 

[FirstDownB]

they've played a good schedule and performed well. They currently rate #5 in RPI, #6 in BPI, #8 in Sagarin, #12 in AP poll, and #12 in Coaches poll.

The better question is why wouldn't they be also rated highly in KenPom?

The opinion of some was that their schedule was lacking.

I agree that it is a good schedule. Not because it is top heavy, but because instead of a lot of 200 and 300 level teams it has a lot of top 200. Which is a combination of smart and lucky scheduling.

I like the KenPom metric and it is one of the best predictors, but I do wonder how truly accurate it is on the spreads of highly mismatched games. Those matchups are so volitile, as you don't know how many minutes starters are going to play or how long the gas pedal will be pushed. Per my example of Purdue's opponents above, there wasn't much difference in scoring margin against the "good cupcakes" and "bad cupcakes". I suspect KenPom would have predicted a larger disparity between the groups, built on data largely from games where they played more similarly matched opponents.

 

[blockm2]

I like the KenPom metric and it is one of the best predictors, but I do wonder how truly accurate it is on the spreads of highly mismatched games. Those matchups are so volitile, as you don't know how many minutes starters are going to play or how long the gas pedal will be pushed. Per my example of Purdue's opponents above, there wasn't much difference in scoring margin against the "good cupcakes" and "bad cupcakes". I suspect KenPom would have predicted a larger disparity between the groups, built on data largely from games where they played more similarly matched opponents.

It's pretty accurate even on massive spread games. For a guess of what point spread would look like, the #200 team on KenPom currently rates as about 9 points better than the #300 team per 100 possessions. In a 65-70 possession game, that would be a difference of about 6 points.

Yes, outcomes in mismatch games can be volatile depending on how long a coach plays his starters and what not, but KenPom doesn't systematically over or under estimate the point spreads for those games to any significant degree.