really;842469507 said:
I'm curious how well Sagarin works in predicting tournament games (especially when such different conferences are matched up).... Definitely something off if UW is still predicted to beat Cal 
It's just a computer algorithm, or actually a couple different ones. I think it's better than the RPI since it does not throw away key pieces of information (margin of victory, opponents' margins of victory over their opponents, and so on). The overall Sagarin rating aggregates 3 ratings, the Predictor, another rating that uses margin of victory called 'Golden Mean', and finally ELO-CHESS, which like RPI does not take margin of victory into account, only wins and losses (home and away), opponents' wins and losses, and so on.
Both the overall Sagarin rating and the Predictor are for judging aggregate accomplishment over the course of a season, just like RPI, except again RPI throws away margin of victory. It's probably good for predicting probable aggregate results if Cal were playing a whole bunch of peer-level teams. As you would expect, it's not so good at predicting individual games or else you could apply it blindly and make serious bank in Vegas (it's interesting though that Vegas lines, which try to equalize odds, often are fairly close to Sagarin Predictor differentials plus home-field advantage). And naturally, it does
not take into account trends, injuries, good or bad match-ups, team morale and confidence. So Cal's rating includes all those games where Range was out or recovering. Blowout losses, like the one at Kansas, or even the ones against good teams like at home against ASU and on the road against Oregon State on their senior night really hurt the Predictor rating especially.
For the men, I think Ken Pomeroy's ratings might be even better than Sagarin, breaking down offensive and defensive efficiency. Not sure if Pomeroy has a rating for the women?
