ranking predictions
- Thread starter cyclones12321
- Start date
It's cute you think you need to educate me, but I'm quite aware of this. A lot of the elements in Kenpom seem very arbitrary to me in terms of actually ranking teams according to the merit of who beat whom. Understandable in terms of predicting the score of a specific game, analyzing trends, or developing coaching strategies, but arbitrary in ranking teams according to their win/loss performance against their schedules.
That's old thinking. If we are going to be a great program we better be able to handle being rated like one.
So why did you slight KenPom for moving teams up and down when they don't play? (BTW, ISU is up 7 now since BYU, even though they were only up 4 the day after.) Every other rating system would work the same way if they polled/computed everyday. Also, in another thread, you said KenPom was discounting ISU's ranking because "we score too much"??? That makes no sense because total game score isn't including in the rankings and tempo is removed.
There's nothing "arbitrary" to KenPom's ranking, at least once preseason ratings are removed... it's a simple equation that estimates winning percentage against an average DI team: Pyth = (AdjOE)^p/[(AdjOE)^p*(AdjDE)^p] (p = 11.5, comes from best fit using past data). Besides that, I don't know what your problem is with KenPom is. (This is more for others in the thread to note than you personally) Like you say, it is supposed to be a system for ranking teams based on who *should* beat whom on a neutral court (i.e. what teams are actually better than others), not a system of "win and you go up, lose and you go down". The same is true for Sagarin Predictor. RPI and Sagarin ELO, on the other hand, are win/loss/SOS based (kind of like coaches and media polls, except without the bias towards winning/losing streaks).
The way I would put it is that a team's schedule is somewhat arbitrary. You get a little information about who's better by looking at win/loss record, and then a little more by using W/L to determine SOS, then weighting W/L's against that. But you get even more info when you zoom in on the data available and look at points per possession (PPP), and even more when weighting that against PPP-determined SOS. No system is perfect (none will ever predict games with 100% certainty), but understanding how a system works has to come before pointing out its actual flaws.
Michigan had a nice win and moved up in both Sagarin and Kenpom. In Sagarin that win boosted us too (5 spots), in KenPom we dropped two spots on the same day. It's my understanding that Kenpom has things built in about tempo and such, I don't feel that has any place in a ranking system, although I can see how it would be useful for predicting games and analysis.
KenPom displays tempo as a diagnostic, but it's not included in (actually, it's removed from) rankings. Tempo is used for score predictions, and thus is part of W/L projections, but those predictions/projections aren't part of the ranking system.
Without pouring over every box score from the last few days, it's hard to pinpoint why a team moves up/down in a day in any computer ranking. Michigan isn't the only team that ISU's O/D ratings are based on, and vice-versa. But for an example, Michigan is now down one (13 from 12) from their OT win over FSU (ranked 33). It looks like KP predicted a Mich win, 73-70 with a tempo of 70 possessions, so 1.043 PPP to 1.000 PPP. The game wound up being 82-80 with tempo 71 (including OT), so PPP were actually 1.155 - 1.127. This means that, following the game, Michigan's ORtg probably went up because 1.15 > 1.04, but their DRtg probably went down (1.13 > 1.00). How that all plays out over time with all the other games that have gone on (because these things all get recalculated in retrospect), well, you gotta have a computer program (KenPom) for that
KenPom displays tempo as a diagnostic, but it's not included in (actually, it's removed from) rankings. Tempo is used for score predictions, and thus is part of W/L projections, but those predictions/projections aren't part of the ranking system.
Without pouring over every box score from the last few days, it's hard to pinpoint why a team moves up/down in a day in any computer ranking. Michigan isn't the only team that ISU's O/D ratings are based on, and vice-versa. But for an example, Michigan is now down one (13 from 12) from their OT win over FSU (ranked 33). It looks like KP predicted a Mich win, 73-70 with a tempo of 70 possessions, so 1.043 PPP to 1.000 PPP. The game wound up being 82-80 with tempo 71 (including OT), so PPP were actually 1.155 - 1.127. This means that, following the game, Michigan's ORtg probably went up because 1.15 > 1.04, but their DRtg probably went down (1.13 > 1.00). How that all plays out over time with all the other games that have gone on (because these things all get recalculated in retrospect), well, you gotta have a computer program (KenPom) for that
I didn't realize the tempo stuff wasn't included in the rankings (and maybe it was included last time I read up on it). I assumed it was since Wisconsin is often 10-15 spots higher in Kenpom than they are in other computer rankings even late in the season. I follow lots of these in football and basketball because they frankly make more sense to me than media polls and especially coaches polls. I'm especially a big fan of an average of computer formulas like the 1/3 of the BCS.
I just don't get the over the top love for the Kenpom rankings. I get that he's much better at social media/blogging than most of them and there's a ton more to his site than just rankings, but at the end of the day I find myself nodding in agreement with Sagrain and even RPI sometimes toward the end of the season than Kenpom (or at least that I don't get the raving about it when it comes to ranking teams according to the quality of their wins).
Can we have this debate in a few more weeks? There's no way there's enough data for any computer rankings to mean a damn thing yet.
I think some of them like Sagarin and Kenpom are just as meaningful as any voter polls. Obviously the ones that have MSU in the 40s or 50s are pretty meaningless. But how can a voter rank team like Iowa right now any better than a computer? They pass the eye test, but they've played almost the dead last strength of schedule? A computer can do just as well in a lot of situations early in a season as voters, just my opinion.
Last edited:
Mostly, I just find his (A) easier to understand (methodology is all spelt out in his blog) and (B) to have lots of added value over other systems. Difficult to find tempo, rebounding percentages, eFG%, etc. on other sites (at least for teams not in the Big 12, thanks to khaal's Cyclonepedia). The social stuff is indeed a bonus.
(And no, DurangoCy, I'm not Ken, wish I had his job though!)
Last edited:
Iowa State may move up a few North Carolina will get major mileage out of that Louisville win. Voters will get Belmont-amnesia and move UNC up at least seven spots.
VCU will drop, but possibly not below ISU. New Mexico will fall below Iowa State.
UMass is going to enter top 25 if it beats Clemson tonight. Florida State could creep into the picture (won't pass us).
Interesting case is Michigan. If U-M loses to Charlotte, that waters down ISU's win ... Iowa State would jump Michigan. With a win, it may hold steady (depending on how much value there is in beating FSU) but that would be better resume-wise.
Highest I see in AP is 17th
In coaches poll, ISU was 27th last week, and BYU had 1 vote, few teams at the fringe lost — 24th.
VCU will drop, but possibly not below ISU. New Mexico will fall below Iowa State.
UMass is going to enter top 25 if it beats Clemson tonight. Florida State could creep into the picture (won't pass us).
Interesting case is Michigan. If U-M loses to Charlotte, that waters down ISU's win ... Iowa State would jump Michigan. With a win, it may hold steady (depending on how much value there is in beating FSU) but that would be better resume-wise.
Highest I see in AP is 17th
In coaches poll, ISU was 27th last week, and BYU had 1 vote, few teams at the fringe lost — 24th.
Uh Oh, Michigan is losing to Charlotte. If that holds, will that take some of the luster off of that win?
If that was our only quality win of the last week. And like last week a lot of voters will have already turned in their ballots.
