Ken Pomeroy’s Game Estimates Have Been Pretty Good For the NCAA Tournament
Ken Pomeroy, the meteorologist who also runs the popular college basketball website, kenpom.com, is a go-to resource for me daily during the college basketball season. He gets lots of praise, and also plenty of criticism when his rating system ranks a team differently than public perception. After every instance where a team lost that was ranked highly in kenpom, or won that was lightly regarded, I saw tweets, notably from Mike DeCourcy, about the failure.
Obviously, pointing out one or two examples is cherry picking. To evaluate, we need to compare the entire body. The other misconception is that if a rating system says teams have a 55% chance, and then those teams win every game, it is a success. If you say that a team has a 55% chance, then it should lose 45% of the time, or the rating is suspect.
Pomeroy publishes his game data in a “FanMatch” page (behind a pay wall) that lists game probability and projected score. Here’s a screen grab of one from the first day of the tournament (I’m sure he won’t mind me posting this subscriber content).
So, how did the Pomeroy Ratings do leading up to the Final Four? I resolved to track this no matter the outcome, and report it. California looked horrible in the first round, Missouri lost to Norfolk State in a huge upset when Pomeroy had Norfolk rated as a 16-seed, Wichita State was 10th in his rankings, but lost in the first game and Belmont was again rated too high and lost in the first round.
There have been 64 tournament games. With all of those results, the Pomeroy favorites were projected to collectively win 69.7% of the games, which converts to 44.6 wins for the favorites in the tournament games until now. The favorites actually won 45 games. Pretty good for government work. Now, if we can get our weathermen to be as accurate.
If you are curious, today’s average projected outcomes are Kentucky over Louisville (75%) and Ohio State over Kansas (60%).
[US Presswire]

- Bob Stoops: Home Was Burglarized, Had Car Stolen
- Muhammed “King Mo” Lawal’s KO of the Year Candidate Against Seth Petruzelli at Bellator 96
- Break-Up Song About Players Leaving Via Free Agency Should Make Fans Cry
- Fight Master: Bellator MMA: Freshening Up a Tired Format
- John Wall Tattoos Are of the Raleigh Skyline and ‘The Great Wall’

- PL StabbinKabin on Muhammed "King Mo" Lawal's KO of the Year Candidate Against Seth Petruzelli at Bellator 96
- PL StabbinKabin on Neymar's Silky Dribbling Schools the Mexican Defense, Sets Up Another Brazilian Goal [Video]
- VladimirCrouton on Bob Stoops: Home Was Burglarized, Had Car Stolen
- KC Resident on Break-Up Song About Players Leaving Via Free Agency Should Make Fans Cry
- A.P. on Break-Up Song About Players Leaving Via Free Agency Should Make Fans Cry
7 Responses to “Ken Pomeroy’s Game Estimates Have Been Pretty Good For the NCAA Tournament”
Leave a Reply
You must be logged in to post a comment.







March 31st, 2012 at 12:47 PM
KenPom overrates Wisconsin because of Ryan Evans’ awesome high-top fade…even Sullinger is overjoyed by it
March 31st, 2012 at 1:00 PM
Ken Pomeroy, the meteorologist
He can’t predict the weather and now I’m supposed to expect he can predict basketball games?
/ignores all data.
March 31st, 2012 at 1:03 PM
Are we getting a baseball post today? Because I just saw that Freddie Sanchez went on the DL again, and I want someplace to commiserate with the Giants fans who are sutkc with Freddie and Zito.
March 31st, 2012 at 1:36 PM
After every instance where a team lost that was ranked highly in kenpom, or won that was lightly regarded, I saw tweets, notably from Mike DeCourcy, on multiple occasions every time a team did not play well that was favored by Ken Pomeroy’s ratings.
What?
March 31st, 2012 at 1:42 PM
What?
Yeah, that was not my finest sentence.
March 31st, 2012 at 2:11 PM
I thought you are supposed to only report it if it shows what you want it to show????
/seriously, good job on that though
How does that compare to if you just went by historical W-L record?
March 31st, 2012 at 2:35 PM
Doesn’t WHICH games were picked correctly matter, a lot?
Say I picked 4 games, with faves of 55%, 75%, 75%, and 95%. If I get any single game wrong, my score in your evaluation is 3 correct vs. 3 expected correct. Great job! But it seems to me that getting the 95% wrong is worse than getting the 55% wrong.
Maybe adding up the predictions of the teams that actually won would work better? Noodling around a bit in Excel, I came up with this, not sure if it is really mathematically sound, but it seems to measure what I am looking for:
AVG(projectedWin% of actual winners)/(predictedWin% of picks)
In my example above…
Getting 75/75/95 right & 55 wrong would give you 97%.
Getting 55/75/75 right & 95 wrong would give you 70%.
If you predicted all at 75% & got one wrong, you would get 83%.
If you predicted all at 100% & got one wrong, you would get 75%.