Re: Power Ratings, Week 13
posted at 12/3/2009 2:28 PM EST
In Response to Re: Power Ratings, Week 13
[QUOTE]Paul does your system make any distinction about how many or few points a team is up at the end of three quarters? If so, how? Just trying to understand your system's logic.
Posted by underdoggg[/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 desires. I claim that each 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 occasionally to try out the rookies. For these reasons I count only the first three quarters of point differential as a statistical signal for future game potential. Yes, blowouts count for more differential.
Currently I consider scoring in the fourth quarter as too much noise to get a useful signal. If I were a pro I might sometimes, but never during blowouts, score the first half of the fourth quarter too, and on top of that I would want to factor in field position at the cutoff time point. For example, a team with first down on the opponent’s one yard line at the cutoff point deserves at least 6 points of credit. This extension into the fourth quarter would help to satisfy the fourth quarter mavens. However, all those gradations would take too much work for me to do. I just use third quarter scores for now.
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. For calculating my point spreads, 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. Most stats connected to this league assume that the September results are just as significant as last week’s results. I don’t. Good ratings start to go stale as teams change. Many teams see season-ending injuries to key players and of course the teams stay that way. Sometimes newbie defenses and offenses start to get better. I have an arbitrary weighting factor that applies a 10% per game stale rate. I want current power rankings for teams, not some old laurels. So, that infamous Tennessee game is fading fast.
6. My spreadsheet then goes to work and finds out what 32 NFL power rankings would best fit the above criteria with minimum variance, over a season full of 100+ games.
So far I like my power rankings. They predict next week’s Vegas point spreads with modest accuracy. More important, I can predict future scores, in a completely different way, in a statistically independent way, which is great.
Note that my stats are completely blind to quite recent injury reports. I often qualify my stats. At this point I tend to combine my independent information with various conventional sports columnist opinions and rankings.
As always, gambling is for people who don’t understand mathematics.