Dear Ken Pom


(Matt Jackson) #1

Throw out your formula…

15-0 and still can’t crack your top 30?

10-4-1 against the Vegas Spread. Among top teams, only Gonzaga is better.

Moved up to #7 in RPI.

Probably still #4 in Net ranking.

Should move into top 15 in polls I hope.


#2

Right on!


(Jimmy Morris) #3

It’s obvious KenPom and Sagarin both put way too much weight in their biased preseason rankings that are purely based on flawed recruiting rankings.

How dumb is this part?
Virginia Tech 13-1, 299 AdjSOS, 327 NCSOS, ranked 8th
Texas Tech 13-1, 304 AdjSOS, 340 NCSOS ranked 9th
NC State 13-1, 350 AdjSOS, 351 NCSOS ranked 18th
Houston 15-0, 295 AdjSOS, 316 NCSOS ranked 31st


#4

Seldom right and wrong again. Preseason rankings are phased out by this point.

The difference between UH and those teams with similar schedules and similar records?

Efficiency.

Average scoring margin:

NC St +24.6
VIrginia Tech +24.2
Texas Tech +23.1
Houston +16.4


(Jimmy Morris) #5

Hey genius, you worked hard to find something that separates those teams from Houston but forgot one thing, how it ranks the others. If scoring margin is the big difference, then tell me why Texas Tech is ranked 9 spots ahead of NC State now? Logic not your strong point?


#6

BPI, the other primary credible analytical tool has us at 27.

No model is perfect and these might be overweighting some early season results like not blowing out Rice and barely beating UTRGV, each at home. We also may be better than the model indicates because our defense and great coaching put us in position to win close games more than statistical analysis would generally project. Also, Oregon dropping in the rankings after Bol Bol’s injury may unfairly hurt us. And Jarreau’s return and emergence hasn’t been fully accounted for yet, possibly. However, almost every pretty good team that we beat at home happened by a margin suggesting that we probably would have lost had that game been on the road.

If you plug us vs. Memphis right now at the Fertitta Center into Kenpom’s matchup predictor it has us winning by 11. We won by 13. I don’t think he needs to re-evaluate the model based on that game at least.


(Jimmy Morris) #7

Plugging in numbers after a game, includes the game that was already played in the numbers. WarrenNolan.com had us as a 11 point victory before the game was played. Warren has us as a 7 point favorite over Temple. Where does Kenpom put us before the Temple game?

FYI this is why NET ranking max the scoring margin by 10. Every time Houston has had a 20 point margin, we have gone 11, 12 and 13 deep on the roster. We are now a legitimate 10 deep squad because we played guys that weren’t ready for primetime to get them ready for primetime. Ranking teams lower because of margin of victory in November games against terrible teams is deeply flawed.


#8

That’s true about including data after the fact. However, both us and Memphis barely moved in his overall rankings after the game so it’s really unlikely that anything would have changed much. Maybe the model would have had us like -10.

Kenpom has us as a 1 point favorite at Temple. I guess Kenpom thinks Temple is much better than Warren Nolan.

I agree that there are flaws with overrating MOV in blowouts. But there are also major flaws in capping MOV at 10. Regardless of the why, beating Rice by 40 tells me that a team is probably better than someone who beat Rice by 10 if those are the only two data points we have. Ideally MOV would gradually start to phase out and mean less past certain levels. I don’t know if Kenpom does this.

And sure, injuries (Jarreau), suspensions (Davis first half against NW State), poor officiating in a given game, etc. aren’t going to be properly accounted for in almost any model that can realistically be put together and updated frequently, until we get quantum computing maybe. But that stuff evens out over time and it doesn’t seem like we’re moving at all in Kenpom, which would seem to indicate most of our results have been in line with what his model expects.


(Jimmy Morris) #9

But that isn’t the only data points right? Location and time of the year are other data points that can easily be applied. If we beat Rice once by 10 points in November at TSU and 4 weeks later beat Rice by 25, late December in FC, which would you weigh more going into a home game against Memphis in FC?


#10

The latter of course, and I think Kenpom does factor in recency of results to an extent. But overweighting recency bias is also a common pitfall in standard evaluation.

Also, you indirectly bring up a good point that maybe our home court advantage really wasn’t one in the first four games at TSU (though going undefeated there last year probably indicates otherwise) and Kenpom isn’t controlling for that. Also, maybe lack of familiarity at FC didn’t give us as much of a home court advantage early at Fertitta (though that place was rocking against Oregon and LSU).

One thing I meant to add is that NET doesn’t cap MOV at 10 because it’s better in terms of statistical analysis. The purpose is to not encourage teams to run up the score and to encourage better sportsmanship.

Like I said, Kenpom isn’t perfect and it may be unfairly dinging us a little. But if you look at our results against other teams we’ve played and where they are ranked it seems to make some sense intuitively. If we beat Temple by double figures on Wednesday we’ll likely move up a few spots. If we beat them by 1 or 2 our ranking won’t change much.

Sampson himself has very recently said this team is good and gritty, but isn’t an extremely good team. However, it has a shot to become one by tourney time if guys like Hinton, Alley and White can improve a lot in the next couple of months. That could just be the message he wants the team to hear, but Sampson generally tells it like it is.


#11

Except he’s not wrong, you jumped the gun by a couple of weeks. Preseason rankings have some weight until l as later this month.


#12

I wouldn’t exactly called him a genius, more like someone with too much time in their hands :joy::joy::joy:
[@CougarRed]
:mask::point_up_2::point_up_2::point_up_2::point_up_2:


#13

I have watched 85+% of the State and UH minutes played this season and I think that would be a pick em type game on a neutral floor. A lot of State’s point margin is simply style of play. Keatts has 10 players. They all play. They’ve all scored in double figures this year. He presses the committee had multiple metrics on each team sheet and I’m sure they all have the ones they prefer so it will be interesting to watch the teams settle into thei slots in each of these systems as the body of work expands. I have a lot more faith that UH will hold up well than I do State. That’s equal parts schedule and style of play. I still have doubts about all pressure all the time and I think Sampson’s style produces more consistent results.


#14

My bad. But if it’s not completely phased out by now by KenPom, preseason ratings are barely weighted. With each game, they are less and less important. They are phased out in other ranking systems by now.

And so he’s certainly wrong to suggest that preseason rankings are the main reason Houston is rated behind those other schools. I identified the main reason above.

Heck, Texas Tech was 25th in the preseason, NC State 36th, VaTech 22nd. It’s not like those teams started out in the KenPom Top 10. Houston was about 40th if I recall.

The fact is, all of the efficiency rankings have Houston about 30th. Not just KenPom.

KenPom = 31st
T-Rank = 29th
BPI = 27th
Sagarin = 30th
Dokter = 33rd
TeamRankings = 33rd
StatFox = 25th

It’s a conspiracy.


#15

The main takeaway is that all methods of judging team strength this early are flawed


(Patrick) #16

I’ll throw this into the conversation: Massey Composite has us at 16

https://www.masseyratings.com/cb/compare.htm

Rating services that they track have us anywhere from 3 to 39.


(Jimmy Morris) #17

Funny how you left out RPI, NET and 10 other ratings/polls that have us ranked way higher than the ones you listed.


(Patrick) #18

AAC NET Rankings


#19

NCAA Team Sheets Released

Note that RPI is gone
KPI Had the Best Correlations Last Year

Enjoy these two links


#20

The most obvious quick take-away is that according to Net our strength of schedule is in the 80’s not 300.
And our average victory is a higher Net ranking than Virginia’s.

Thanks for the chance to waste a whole day, Pack😉