Re: Colts average team..
posted at 11/15/2009 2:42 PM EST
In Response to Re: Colts average team..
[QUOTE]In Response to Re: Colts average team.. : How do you generate your "moment in season" ratings? Do you average the last few games and remove points for potential loss of strength via injury? I think it is a far better indicator when matching teams on a week to week basis but haven't seen a formula or method. Gaspar, or one of the other guys who took over Reiss' blog has an interesting method that he posted (just the results, not the calculations) last week.
Posted by EnochRoot[/QUOTE]
My completely homemade stat is the number of victories that a team would achieve if the season restarted tomorrow, and if teams played each other at random. It works on the following principles.
1. I count point differential because it’s exactly what each coach wants. A team plays its most predictive football for three quarters. In the fourth quarter teams take on strategies not to maximize points, but to seal the win, to gamble for the win or to try out the rookies (in the Tennessee game). For these reasons I count only the first three quarters of point differential as a statistical signal for future game potential. I consider scoring in the fourth quarter as too much noise to get a useful signal.
2. A team that dominates time of possession statistically does better in the fourth quarter. Whole-game time of possession (excluding overtime) is independent of point differential. I arbitrarily give 1 point for each 2.5 minutes of possession over 30.
3. A home team should be three points better.
4. I arbitrarily assume that a team with a strength of 9-7 will be 1.5 points better than a team with a strength of 8-8.
5. Good ratings start to go stale as teams change. Many teams see season-ending injuries and of course they stay that way. Sometimes newbie defenses and offenses (such as the Patriots) start to get better. I have an arbitrary weighting factor that applies a 10% per game stale rate.
6. Note that my stats are completely blind to quite recent injury reports. I often qualify my stats. In this case, Julian Edelman, Matt Light and Dan Koppen all took the trip out to Indianapolis, and that's new, while on the Indy side, their favorite remaining cornerback is now out, and that's new.
7. My spreadsheet then goes to work and finds out what 32 NFL power rankings would best fit the above criteria with minimum variance. I can then predict future scores, probably not as well as the Vegas sharks with millions of dollars of research, but in a completely different way, which is great.
So far I like my power rankings a lot better than, say, Mr. Breer's rankings in the Globe, but I have almost zero signal against the Vegas spread.