Sagarin Based Bowl Pick'Em Sheet
Sometime a couple of years ago, someone on the board created a Bowl Pick'em spreadsheet that used Sagarin ratings to create a "Watchability index" that looked for games where Sagarin predicted close games and used this budget his time over the holidays, and also to help him pick winners. I haven't seen anyone post this yet this year, so I figured I'd go ahead and make one up.
Here is a link to the spreadsheet. It is setup to work with the Yahoo Sports system, so the games are sorted in order of most confident picks (Arizona State over Navy) to least confident (Northwestern of Mississippi State). There's a "PREDICTOR Pick" column which indicates which team the PREDICTOR model would pick to win the game (Sagarin regards PREDICTOR as the best model of game outcomes). If you prefer, you can also pick by RATING and ELO_CHESS differences using the appropriate columns in the sheet. Enjoy.
December 10th, 2012 at 4:34 PM ^
Interesting that NW over Mississippi State is the least confident pick. I may be biased, but I think NW will beat them pretty easily. Mississippi State hasn't beaten anybody, and while NW hasn't really either, they really aren't that far from being 12-0.
EDIT: D'oh, it has Mississippi State over NW. But still, I think NW should take this game.
December 10th, 2012 at 4:18 PM ^
If I'm reading this spreadsheet correctly, then no B1G team is favored to win their bowl game. Should be an interesting bowl season!
December 10th, 2012 at 4:23 PM ^
Northwestern and....Rutgers, I guess. The "Underdog" column is based on Yahoo's classifcation, not Sagarin's.
December 10th, 2012 at 4:25 PM ^
Ah yes, thank you...brain fried from a long day in lab.
December 10th, 2012 at 4:41 PM ^
December 10th, 2012 at 4:52 PM ^
The "PREDICTOR Pick" column on the far right tells you who Sagarin's PREDICTOR model says should win the game, and the "PREDICTOR dIfference" column tells you by how much. Games at the top of the list should be blowouts, the games at the bottom should be close.
December 10th, 2012 at 5:02 PM ^
December 10th, 2012 at 5:47 PM ^
Basically, using the Sagarin predictors, the Pac-12, Big XII and SEC are estimated to win six bowls each, the SunBelt could potentially win four, and the WAC, MWC and Big East are predicted to walk away with two each, if the model ends up to be totally spot on. The MAC, C-USA, ACC and the Big Ten are favored in only one. Then, of course, there is BYU. Another action-packed bowl season ahead, from the sound of it.
December 10th, 2012 at 8:23 PM ^
expected wins <> # of bowls favored in.
If you have 4 teams each with a 48% chance of winning, your expected wins are just under two but you aren't favored in any of them.
December 10th, 2012 at 8:28 PM ^
is a mug's game for the Big Ten. The problem is that we've got less to complain about this year than many years, and our teams are so bad we still may lose big. Sigh.
December 11th, 2012 at 9:28 AM ^
I can't find the link. I lost even before I started.
December 11th, 2012 at 11:56 AM ^
It's right there on the here
December 11th, 2012 at 2:41 PM ^
To measure the luckiness (or surprise value) of a season record, you might want to consider the standard deviation as well as the mean number of (expected) wins. If a team is good and expects to win 90% of its games, the standard deviation will be much smaller than if a team is mediocre and expects to win only 50%. So, winning one game more than expected will signify greater "luck" (or surprise us more) for a very good team than for a mediocre one.*
*suppose then that you measure such "luck" as the deviation from the expected number of wins divided by the std dev. Say then number of games in the season is n=16.. Suppose that all the games were against equivalent opponents, and the win probability for each game were p=7/8 for the very good team vs p=1/2 for the mediocre team. Then the std dev (= (n*p*(1-p))^(.5)) would be about 50% lower in the former case, resulting in a much greater "luck" value for winning each additional game beyond expectations. .
Of course, if the very good team goes 16-0 (vs 12-4 expected) and the mediocre one goes 12-4 (vs 8-8 expected), you might also want to revise your estimates of p. That would change the numbers a bit, but under any reasonable assumptions, still result in a greater "luck" value for the very good team.