SEC and Efficiency Ratings (i.e. KenPom and NET)

[nelsonmuntz]

Post in thread 'Inter-conference records among the majors'

Jan 5, 2026

As the efficiency ratings and RPI have become more seasoned with additional games, the results are fairly consistent across NET, Torvik, KenPom and RPI. The Big 10 and Big 12 are essentially tied, followed by a gap then the ACC, followed by a smaller gap, then the Big East, followed by a huge gap then the MWC, WCC, A10 and everyone else, with the mid majors clumped pretty close. The efficiency ratings and RPI do not meaningfully disagree on the vast majority of leagues.

Then there is the SEC. The efficiency ratings, and particularly KenPom, believe this year’s SEC is one of the...

• nelsonmuntz

• 

I started a thread three months ago about the interconference records compared to the power ratings, and questioned why the SEC was so highly rated when the actual game results would indicate a much less impressive conference.

KenPom had the SEC at +19.46 going into the tournament, making it the 5th best conference of all time for KenPom, going back to the late 90's.

Well, after the first weekend, the SEC only got 4 of 10 teams into the Sweet 16, but it is worse if you dig in.

Sweet 16 SEC teams:

Alabama - impressive win vs. Texas Tech, but they literally had a major drug dealer on their roster. If there was ever a reason for a team to be forfeited out of the tournament, this would be one. NL yet for next round.
Arkansas - Beat a 13 and 14 seed, and the High Point game was close. +8.5 in the next round.
Texas - Beat BYU and Gonzaga, who were both missing key players. +7.5 in the next round
Tennessee - beat one of the weakest 3 seeds in recent history. +4.5 in the next round.

There is a reasonably high chance that the SEC exits the tournament in the next round.

SEC Teams who didn't make it to the Sweet 16:

Florida - lost to a team 8 slots below it
Vanderbilt - respectable loss
Texas A&M - obliterated by Houston
Kentucky - needed a miracle to beat Santa Clara, and was easily handled by an Iowa State team missing one of its best players
Georgia - demolished by St. Louis in first round
Missouri - first round exit in game that wasn't close

The SEC is 3-6 against the major conferences in the tournament, with two of the wins close and 4 of the losses blowouts.

Does this look like the 5th best conference of all time? There were several homages to KenPom's brilliance as a predictive model. Did KenPom predict that the 5th best conference of all time would fall flat on its freaking face? My theory Is that the SEC was overrated in the efficiency ratings because it ran up the scores against the bad teams on its schedules. I am open to other explanations for how the efficiency ratings systematically overrated SEC teams.

 

[FightOnCT]

I think it’s as simple as there being a lot of solid but not elite teams. I do agree that a lot of these analytical numbers are being jacked up but because teams have gotten better at smoking inferior competition.

Was never sold on Florida, their KenPom numbers were sky high but they never looked particularly extraordinary compared to other good teams in other years despite the metrics implying that.

 

[HooperScooper]

 

[superjohn]

Why do you do this to yourself?

 

[auror]

Of the 10 teams from the conference that made the tournament, 8 met or exceeded their seed expectations, 2 were upset early. And that includes an 8 losing to a 9 as an "upset".

 

[formerlurker]

Something tells me you had your bachelor party at Costco.

 

[Dogfish]

Isn't this always about football and awarding conferences for such?

 

[nelsonmuntz]

Of the 10 teams from the conference that made the tournament, 8 met or exceeded their seed expectations, 2 were upset early. And that includes an 8 losing to a 9 as an "upset".

I forgot about the season ending injury to Toppin from Texas Tech until after I made the first post. Even Alabama‘s win over TTU came with a big asterisk.

The fifth best conference of all time needed 3 very friendly matchups against teams missing key players to even get 2 of its teams into the Seeet 16. It isn’t just losing based on seed, the SEC looked terrible doing it. I would argue that several of the surviving teams were helped by generous seeding due to high efficiency ratings that they did not deserve.

Why is it so hard for some people to admit that KenPom, which has an obvious flaw that teams are rewarded for running up the score, can be gamed by teams or a conference systematically running up the scores? Given that I pointed out this problem in December, one of two things must be true: 1) I am a genius who can see the future, or 2) this is an obvious problem with efficiency ratings and anyone refusing to acknowledge it is just being stubborn.

 

[Old Man UConn]

The ACC is not overrated, they just suck. Big East is 4-1.

 

[Dibbie]

I’m just glad FL lost as they cost me money, trying to dunk the ball (leading to a fast break) up 60 with 30 seconds to go in their first game. I prefer Hurleys respectful let the shot clock expire approach. Golden is a tool.

 

[7774]

My favorite part of this is how you're using open lines, which just mirror Kempom's point spreads, as evidence that Kempom is bad.

 

[John]

Something tells me you had your bachelor party at Costco.

The $1.50 hot dogs come with a free soda

 

[jfreeds]

Post in thread 'Inter-conference records among the majors'

Jan 5, 2026

As the efficiency ratings and RPI have become more seasoned with additional games, the results are fairly consistent across NET, Torvik, KenPom and RPI. The Big 10 and Big 12 are essentially tied, followed by a gap then the ACC, followed by a smaller gap, then the Big East, followed by a huge gap then the MWC, WCC, A10 and everyone else, with the mid majors clumped pretty close. The efficiency ratings and RPI do not meaningfully disagree on the vast majority of leagues.

Then there is the SEC. The efficiency ratings, and particularly KenPom, believe this year’s SEC is one of the...

• nelsonmuntz

• 

I started a thread three months ago about the interconference records compared to the power ratings, and questioned why the SEC was so highly rated when the actual game results would indicate a much less impressive conference.

KenPom had the SEC at +19.46 going into the tournament, making it the 5th best conference of all time for KenPom, going back to the late 90's.

Well, after the first weekend, the SEC only got 4 of 10 teams into the Sweet 16, but it is worse if you dig in.

Sweet 16 SEC teams:

Alabama - impressive win vs. Texas Tech, but they literally had a major drug dealer on their roster. If there was ever a reason for a team to be forfeited out of the tournament, this would be one. NL yet for next round.
Arkansas - Beat a 13 and 14 seed, and the High Point game was close. +8.5 in the next round.
Texas - Beat BYU and Gonzaga, who were both missing key players. +7.5 in the next round
Tennessee - beat one of the weakest 3 seeds in recent history. +4.5 in the next round.

There is a reasonably high chance that the SEC exits the tournament in the next round.

SEC Teams who didn't make it to the Sweet 16:

Florida - lost to a team 8 slots below it
Vanderbilt - respectable loss
Texas A&M - obliterated by Houston
Kentucky - needed a miracle to beat Santa Clara, and was easily handled by an Iowa State team missing one of its best players
Georgia - demolished by St. Louis in first round
Missouri - first round exit in game that wasn't close

The SEC is 3-6 against the major conferences in the tournament, with two of the wins close and 4 of the losses blowouts.

Does this look like the 5th best conference of all time? There were several homages to KenPom's brilliance as a predictive model. Did KenPom predict that the 5th best conference of all time would fall flat on its freaking face? My theory Is that the SEC was overrated in the efficiency ratings because it ran up the scores against the bad teams on its schedules. I am open to other explanations for how the efficiency ratings systematically overrated SEC teams.

You are weighing a small sample size over a large sample size to support your bias and desire to "prove" the SEC is overrated. When you use a small sample size like this it is much harder to weed out injuries, bad calls, crowd advantages, off-nights shooting the ball, bad match ups etc. When you take a large sample size you can start to apply the law of averages. If the tournament started again next week all these spreads would be the same. Florida would still be a 10-point favorite over Iowa.

Were the efficiency ratings wrong last year when the SEC dominated?

 

[FightOnCT]

I forgot about the season ending injury to Toppin from Texas Tech until after I made the first post. Even Alabama‘s win over TTU came with a big asterisk.

The fifth best conference of all time needed 3 very friendly matchups against teams missing key players to even get 2 of its teams into the Seeet 16. It isn’t just losing based on seed, the SEC looked terrible doing it. I would argue that several of the surviving teams were helped by generous seeding due to high efficiency ratings that they did not deserve.

Why is it so hard for some people to admit that KenPom, which has an obvious flaw that teams are rewarded for running up the score, can be gamed by teams or a conference systematically running up the scores? Given that I pointed out this problem in December, one of two things must be true: 1) I am a genius who can see the future, or 2) this is an obvious problem with efficiency ratings and anyone refusing to acknowledge it is just being stubborn.

What is your suggestion to fixing the metrics then?

 

[nelsonmuntz]

What is your suggestion to fixing the metrics then?

It would be pretty simple to flatten out the impact of 30+ and 40+ point wins in the efficiency ratings like KenPom and NET. Beating Wagner by 40 points should not be meaningfully different in a computer model from beating Wagner by 30 points. The problem with doing that is KenPom has figured out that gamblers are using his model, and may be the biggest customers. If you are using the models for betting, then the willingness of a low rent coach like Golden to run up the score is very relevant because you want to know if Florida will cover a big spread. But if a selection committee is using the model to seed and select the field, then those extreme margins should have a rapidly declining impact on a team's efficiency rating, and if they had done that this season, we would have seen that Florida was good, not great this season.

A simpler metric like RPI had all 5 major conferences fairly close to each other, whereas KenPom had huge differences, particularly with the SEC. The reality was that there was not a huge gap between the leagues and the teams in the leagues, just a huge perception gap because some leagues were systematically gaming the efficiency ratings.

I am fine using efficiency ratings for tournament field selection and seeding if they do not reward running up the score, because this year, those non-competitive games against non-competitive teams had a big impact on the final ratings.

 

[businesslawyer]

It would be pretty simple to flatten out the impact of 30+ and 40+ point wins in the efficiency ratings like KenPom and NET. Beating Wagner by 40 points should not be meaningfully different in a computer model from beating Wagner by 30 points. The problem with doing that is KenPom has figured out that gamblers are using his model, and may be the biggest customers. If you are using the models for betting, then the willingness of a low rent coach like Golden to run up the score is very relevant because you want to know if Florida will cover a big spread. But if a selection committee is using the model to seed and select the field, then those extreme margins should have a rapidly declining impact on a team's efficiency rating, and if they had done that this season, we would have seen that Florida was good, not great this season.

A simpler metric like RPI had all 5 major conferences fairly close to each other, whereas KenPom had huge differences, particularly with the SEC. The reality was that there was not a huge gap between the leagues and the teams in the leagues, just a huge perception gap because some leagues were systematically gaming the efficiency ratings.

I am fine using efficiency ratings for tournament field selection and seeding if they do not reward running up the score, because this year, those non-competitive games against non-competitive teams had a big impact on the final ratings.

Isn’t the better answer to let the gamblers use their tools (Ken Pom) and have the Committee, which is the only rankings we care about, rely on resume tools (as well as detailed analysis of the records)? They seemed to be doing that this year. That’s why all the coverage and explanation of the WAB.

 

[nelsonmuntz]

Isn’t the better answer to let the gamblers use their tools (Ken Pom) and have the Committee, which is the only rankings we care about, rely on resume tools (as well as detailed analysis of the records)? They seemed to be doing that this year. That’s why all the coverage and explanation of the WAB.

Scoring margin is capped in NET at 10 points, but net efficiency is not capped.

 

[auror]

It would be pretty simple to flatten out the impact of 30+ and 40+ point wins in the efficiency ratings like KenPom and NET. Beating Wagner by 40 points should not be meaningfully different in a computer model from beating Wagner by 30 points. The problem with doing that is KenPom has figured out that gamblers are using his model, and may be the biggest customers. If you are using the models for betting, then the willingness of a low rent coach like Golden to run up the score is very relevant because you want to know if Florida will cover a big spread. But if a selection committee is using the model to seed and select the field, then those extreme margins should have a rapidly declining impact on a team's efficiency rating, and if they had done that this season, we would have seen that Florida was good, not great this season.

For the 10,000th time, he adjusts the impact of the games by the rating disparity of the teams, so gross mismatch games count for less.

 

[nelsonmuntz]

For the 10,000th time, he adjusts the impact of the games by the rating disparity of the teams, so gross mismatch games count for less.

For the 10,001st time, a 40 point win over a 340 ranked team counts for a lot more than a 25 point win over a 339 ranked team. There is an easy solution and the only reason he doesn't do it is because his customer base is gamblers. I am fine with that, but then the same metrics should not be used for selection and seeding.

Also, the NET formula absolutely needs to be publicly available.