2007-2008 Rankings (2/9/08)
Explanation of bigeasthoops.com’s ratings
(The ratings on these pages were generated using data from Ken Pomeroy’s website, where he generously shares his NCAA game log. I also credit Ken for his and others’ approach of rating teams offensively and defensively by calculating how efficiently they produce and defend per possession. Ken saved me lots of time and provided insights that have improved my confidence in my results.)
The ratings are derived through an iterative process that is similar to Newton’s Method. Each team’s offensive and defensive proficiency is estimated — conveniently, their raw averages are used as a first pass — and plugged into the schedule. Based on the guess, expected results for each game that has already occurred are calculated. The differences between the actual and expected results for all played games are computed and an average error is determined for each teams’ ratings. The input is adjusted based on this error and the process is repeated until the errors for all teams are sufficiently small. This method accounts for strength of schedule automatically.
Next, the impact of offensive and defensive proficiency is combined to produce each rating. The ratings, as listed on the site, are scaled around the average points scored per game (about 70). To compare two teams and estimate their point differential in an upcoming game, simply substract one rating from the other and make an adjustment for home court advantage. The advantage is around four points (if you’re worried about a half-point swing, these and many other ratings are not precise enough for you).
Ratings measure the past but are intended to provide insight into future games. For this, variation in performance becomes important. A team with a rating of 90 will occasionally lose to a team with a rating of 60. To assess this chance, the final errors for each game are considered. The errors for past games fall very closely on a Gaussian curve. Based on this observation, standard deviations are used to determine the probability of winning games. The standard deviation is around nine points, which means that a nine-point favorite will win about 84% of the time.
One odd observation from the set of errors is that game tempo does not correlate with the magnitude of errors as might be expected. On one hand, it seems logical that the favorite of a projected 60-40 game is more likely to actually win than a favorite of a projected 100-80 game because the former favorite is scoring much more efficiently (3:2) than the latter (5:4). The data does not agree, however. The underdogs in each game are equally likely to make up 20 points in team quality because the performances in the 60-40 game will be more variable over less possessions. The higher variability makes up for the greater difference in efficiency.
Final 2006-2007 Rankings
Final 2005-2006 Rankings
3 Responses to “BigEastology (2/24)”
It’s nice to see the Big East ranked ahead of the ACC. I’m always pleased to see that we get the respect that we deserve. And respect is difficult when you have so many ESPN commentators out there slurping Duke and Carolina constantly. Don’t worry Rafferty and Bilas, we like you.
Obviously once the tournament was all said and done last March, the SEC turned out to be the nation’s best. But still, as long as the Big East is ranked higher than the ACC, I’m having a good day.
Is it true that the Big East players are allowed 6 fouls per game in inter-conference play instead of the 5 that all other D-1 college players get?
John — that’s not true. That used to be the case, but they ended it a while back (at least three years)?
Care to comment?