Editor’s note: TSSAA contributing statistician and Kingston native Earl Nall has been compiling his computer-generated Prep Performance Ratings for high school football for 17 seasons. With football expanding to six public school classifications this fall, Nall is getting an early jump. The weekly Prep Performance Ratings can be found weekly in Blount Today Sports throughout the season.
I am publishing the preseason ratings early so that fans can begin to get a feel of what teams are in the new football classes for the new classification period of 2009-2012.
I am not a fan of preseason computer ratings for football teams. The obvious reason is that the computer model can only make decisions based on historical data. It cannot look ahead and make predictions. To do so would add subjectivity to the statistical model, which in turn defeats the whole premise of what computer ratings are – the processing of facts.
I feel there is no way that someone should put into a computer system the fact that a school graduated all its starters from last year or that the school got a new coach – because one could only “guess” as to what affect these changes may have on the team.
If one was to take a computer model and add that subjectivity to it, the results would be no different than someone’s opinion and that is not in the spirit of computer ratings.
However, there are three facts that I use related to my statistical model for ranking football teams:
1) If the program has good logic and algorithms in it, the system will get better each week. This is the 17th year I have done the Prep Performance Ratings and historically they start the season at about 70 percent accurate and by the end of the 10th week the worst they have ever done is pick 92 percent of the games. Only three times has a team made it to one of the 80-plus state championship game since the ratings began that wasn’t ranked in the top 10 in the final week 10 ratings.
2) The most important factor in a computer modeling system is the ability to have enough data for the computer to investigate the consistency of a team over time. If programmed correctly, the rating program can identity this consistency and parlay that into a pretty accurate preseason rating. Most teams are consistent over time. You find very few that are 8-2 one season and 2-8 the next.
3) The single most important factor in rating teams is also the most obvious - the margin of victory. The reason is simple. Do you know of a case where a team wins a game 35-7 but looses the statistical battle by the same margin? It may happen but it is a very rare anomaly. However, the computer model must also put a law of diminishing returns on large margin of victories. In my system a team gets the same ratings points whether it wins a game 33-0 or 97-0.
Team - Season opener
1 Alcoa - Bell County, KY (Week 0)
2 Milan - Jackson Christian (Week 0)
3 Camden - McKenzie (Week 0)
4 Lewis Co - Mt. Pleasant (Week 0)
5 A-E - Morristown East (Week 0)
6 G’pasture - Whites Creek (Week 0)
7 CPA - Fort Campbell, KY (Week 0)
8 Tyner - Ooltewah (Week 0)
9 McMinn Central - McMinn County (Week 0)
10 CAK - Knoxville Grace (Week 0)
1 Oakland - Ensworth (Week 0)
2 Ooltewah - Tyner (Week 0)
3 Franklin - BGA (Week 0)
4 Maryville - Alcoa (Week 1)
5 Millington - MUS (Week 0)
6 Farragut - Dobyns Bennett (Week 0)
7 Brentwood - LaVergne (Week 0)
8 Whitehaven - Hamilton (Week 0)
9 Dobyns Bennett - Farragut (Week 0)
10 White Station - Briarcrest (Week 0)
43 William Blount - Sevier County (Week 0)
53 Heritage - Karns (Week 0)
55 Arlington Memphis - Overton (Week 0)