We’re midway through Ivy League play, and everyone has played each other at this point. So this is as good a time as any to debut a ratings system I have come up with. Unlike the RPI, my formula looks at the result beyond the winner and loser. And unlike Sagarin and Pomeroy, I don’t just use the final margin, which is often prone to skewing and can give an inaccurate picture of what the game was actually like. And unlike all three of those ratings systems, my formula looks at individual game results instead of just averaging everything.
I’ll just start off by sharing the ratings, based on the games through Tuesday, February 14.
1. Penn (7-0) .649
2. Cornell (5-3) .533
Harvard (4-4) .533
4. Princeton .521
5. Yale .515
6. Dartmouth .426
7. Brown .425
8. Columbia .420
Now, the explanation:
Each game — worth 1.0 points — is divided up amongst the two participants. To account for the home court advantage, I give 0.2 to the visiting team just for showing up. Then I divide the remaining 0.8 points between the two teams, based on how the game plays out, ignoring any distortion due to garbage time or late desperation fouling. The allocation of points is as follows:
0.8/0.0 if it’s a total wipeout (the bench guys get some minutes)
0.7/0.1 if it’s not particularly close, but not a laugher (not in doubt, but the deep bench gets little or no playing time)
0.6/0.2 if it’s decently close, but not down-to-the-wire close (generally single digits, minus drama)
0.5/0.3 if it’s OT or is decided in the final seconds (the losing team had a chance to tie in the final 30 seconds)
I’ll look at the pace of play and the flow of the games as well, and adjust based on that. For example, Princeton being up 15 points on someone is very different from Harvard being up that same margin, because the Tigers’ games have many fewer possessions in them.
I add up those totals for each team, divide by the number of games, and that’s the Team Factor. I compute the Opponent Factor based on the weighted averages of opponents played, and compute the Opponent-Opponent Factor based on that. These components are then averaged together with weighting — the Team Factor counts twice as much as the other two. This results in the following formula:
50 percent Team Factor
25 percent Opponent Factor
25 percent Opponent Opponent Factor
The net result is a formula in which half is how you played, and the other half is who you played.
In case you’re interested here are the final ratings for 2004-05 Ivy League play:
1. Penn (13-1) .627
2. Harvard (7-7) .507
Princeton (6-8) .507
Yale (7-7) .507
5. Cornell (8-6) .497
6. Dartmouth (7-7) .477
7. Brown (5-9) .453
8. Columbia (3-11) .426