Bad Beats: Sagarin's Predictor Rankings and the Bowl Games

I promised more updates, but that obviously hasn't happened. There are three reasons behind this: first, the million distractions of the holidays got in the way, as tends to happen; second, I've been working on my Evolution of Sport proposal for the upcoming Sloan Sports Analytics Conference, and third, it's harder than you'd think to type with these broken thumbs.

Maybe I should explain.

A coworker of mine is in a weekly football pool: for each NFL game, guess which team will cover the spread to win fame and fortune, etc. The pool keeps going into the playoffs, but since there aren't that many playoff games, he's also required to pick any six college bowl games against the spread. Last year, he told me, he used some good guessing and coin flipping to go 3-3. This year, he wanted to do better.

I decided to try and help. Using Jeff Sagarin's predictor ratings, I calculated what the line should be for each of the 33 remaining bowl games as of Dec. 20. I compared this to the Vegas Insider consensus line to pick a side for each game. In the Poinsettia Bowl, for instance, BYU was a consensus three-point favorite over San Diego State. But since Sagarin's rankings made the Cougars a 1.5-point favorite (i.e., less than the consensus line), I took the Aztecs. Since BYU ended up winning by 17, I took a loss for that game.

Once I'd gone through all of the games, I found 12 games where the discrepancy was greater than three points and told my friend, "Here, try these."

Picking from those games, he went a perfect 0-6. Had he picked purely randomly, the odds against that are something like 64-to-1.

Overall, the method did exactly as well as one would predict for flipping a coin: I went 16-16 with one push over the 33 games for a .500 record. There was no predictive power in the larger discrepancies: my record in all 12 games with a discrepancy over 3 points was 5-7. In all, the breakdown was as follows:

Range	W-L-T	Cover %
0 - 1	5-2-1	.714
1.1 - 2	4-4-0	.500
2.1 - 3	2-3-0	.400
3.1 - 4	3-3-0	.500
4.1 - 5	0-1-0	.000
5.1 - 6	0-1-0	.000
6.1 - 7	1-1-0	.500
7.1 - 8	1-1-0	.500

What might be a little more interesting is the winning percentage by date: this method worked well for the first week or so, with a cover percentage of 78% through the first nine games. But with such a small sample, it's hard to tell whether this method was working well until the rankings became outdated, or whether I just got lucky for the first few games. Here's a plot of the cover percentage over time.

Now, I could make lots of excuses, talking about what a crapshoot the bowls are, about the effects of coaching transitions and a one-month layoff on a team's performance in a glorified exhibition game with nothing but pride on the line, but my thumbs are starting to hurt and these bandages aren't gonna change themselves.