College Basketball Elo Ratings
Hi, this is the homepage for my NCAA college basketball ELO Ratings. Wikipedia has a pretty decent description of the basic methodology behind Elo ratings. Also, the excellent World Football Elo Ratings has a good description of their method, which I have adapted and tried to make more suitable for basketball.
Contact: elobasketball@gmail.com
Along with the rankings themselves, I try to keep up a daily blog wrapup of the games occuring in college basketball. This means looking at a 'Game of the Night' that looks to be a close game between high ranking teams, and a 'Result of the Night' which turns out to be an interesting or important game after all the results are in. I also work on a weekly top 25, just trying to get a wider sense of the overall rankings, and a mid-major ranking for less publicized teams.
This is a new endeavour, so the rankings are likely to have some kinks that I'll need to work out. Notably, there is a really huge effect of road/home at this point, perhaps too large. Also, this system is very 'what have you done for me lately' in that it does not recalculate from the beginning when new results are added.
A key effect this has leads me to issue another reminder, namely that the computer rankings know no names. They only know wins and losses, they cannot tell Duke from Duquesne other than by who the teams have played and beat. This leads to a problem, especially with early season rankings, by which teams that we intellectually know won't be top 25 come March are highly ranked, since we were'nt sure how good the teams they beat were. All teams started identically, so it takes some time for the system to normalize, though it will happen.
A final caveat is that this system does consider margin of error, though I'm not terribly comfortable with this, as it favours teams that push against marginally inferior competition (though I do have a cap), and also score is not always indicative of the difference between two teams. However, there's not much I can do about it. I had considered utilizing Ken Pomeroy's work on tempo-independent offense and defense to compare points-per-possession, rather than pure scoring margin, but I'm not neccessarily convinced it would make enough of a difference to make it worth it.