Tag:football playoffs
Posted on: November 2, 2010 12:30 am
 

Computer Pollsters, and some of their basis . . .

A few of you have made some remarks about the Computer Pollsters, and I have a chapter in my book December Dream . . . Qualifying for the Final BCS Rankins which discusses their methods. Of course you may do the research I did, and come up with your own conclusion. I would suggest you cut and paste this for your long term memory, or just buy my book and get all of the remarks, lol Cool, but these are a few segments from my book for your use and information. . .


The Sagarin Poll

Pre-season polls do provide a starting point that can be used by the computer process, as no one really knows who is going to win all or most of their games each season. . . This basis becomes unusable as teams begin to create a record that is more reliable than the pre-season guessing that formulated the pre-season poll, and then the actual records are applied with a recognizable basis to apply to the actual formula. The pre-season poll is usually a biased slant based on the previous years record and returning experienced players. The first few games and subsequent games add to the actual team's performance.

 AndersonSports by Anderson and Hester

It is apparent that this pollster is straight to the points that go into the computer ranking. This formula also ties the conferences into the rating system. A big plus or minus, depending if your in a conference that wins or loses most of their out of conference schedules. And, whereas Sagarin uses the top 10 and 11 to 30 for comparison, Anderson and Hester uses the top 10 and 11 to 25 for comparison.

Billingsley Dynamics by Richard Billingsley

This pollster has a different point of view of assessing point values to teams than the other polls. The first significant difference is the pre-season polls. Mr. Billingsley does not acquiesce to accepting the pre-season polls, but starts each season based on last year's finishing polls to assess his starting values to each team. It is a simplistic system that gives each team opportunity to gain points, based on the opponents they beat, and the points acquired to date, by that opponent. It's pretty obvious that you gain more points by beating a team with a better win loss record. Strength of the opponent determines how many points can be accumulated with a victory. And as the season goes long, the opponent's point values rise, thus giving more emphasis to a team's most recent victory.

 

Mr. Billingsley further states that unlike most systems who use wins and losses to calculate strength of opponents, while his uses a unique twist by applying the opponent's rank and rating. An undefeated team has a ticket to the top 10 as they receive “full earnings” of their opponent's value. A loss on the other hand, allows for a deduction percentage. As the loss column increases, the handicap grows and the only fix is to beat a higher ranked team.

Mr. Billingsley also puts some slight consideration value on where the game is played and the average fan attendance. A capacity crowd in a 40,000 seat stadium will bring a better value than a 20,000 attendance in a 60,000 seat venue.


The Cooley Matrix by Wesley N. Colley Ph.D.

Mr. Colley states that his Colley Matrix: 1- has no bias towards conference, tradition, history, etc. [and hence, has no pre-season poll]; 2- it is reproducible and one can check the results; 3- uses a minimum of assumptions; 4- uses no ad hoc adjustments; 5- none-the-less adjusts for strength of schedule; 6- ignores runaway scores; and, 7- produces common sense results that can compare well to the human polls.

 

Mr. Colley states many advantages to his computer matrix system. His rankings are based only from the results on the field. He uses no pre-season poll, and all teams start from the same basis, and allows no bias from opinion, past performances, tradition or other possible sources. Strength of schedule has a strong influence on his ranking system. . . Winning margins have no effect on this system, as do game locations and weather factors.

The Massey Ratings by Kenneth Massey

 

[My remarks are based on Mr. Massey's August 15<sup>th</sup> 2000 Theory. I requested permission to add this in the Appendix, and Mr. Massey said, “It is outdated, and doesn't refer to the rating system I submit to the BCS. I have two sets of rankings. The one described in your text [my following remarks] you copied [for Mr. Massey's permission] was used until the BCS mandated that margin of victory couldn't be used. I did not ever post any description of that alogrithm.”]

 

The Massey Rating states what is used, and not used in their computer model as such: based on win-loss outcomes relative to schedule difficulty; early season ratings will fluctuate significantly until a sufficient number of games have been played; teams not connected by a schedule graph are rated as isolated groups; these rankings are used in the BCS; these rankings use the MOV formula; and, margin of victory is not used and ratings do not reflect point differential. 

This system does take into consideration the home field advantage, while disregarding the crowd noise, surface, day or night, or weather conditions. . .  This system does measure the ability to score points, but does not distinguish how the points were acquired. . . The schedule of strength is the only representative of games played and depends on where the game is played. . . . The GOF  [Game Outcome Function]

distinguishes between a 10-0 win and a 50-40 win, as a close high scoring game is likely to have more variance and less likely to be dominated by either team. A low scoring game may indicate a defensive struggle or poor game conditions.

Peter Wolfe

This computer system rates all varsity teams at 4 year colleges that can be connected by mutual opponents. If the team's opponents are not comparable, being a community college, JV team, etc., then they are not counted, but the game location is taken into account. This system also rates teams on a won loss record, and not does not take into consideration run up scoring.


 And lastly, a few of my comments included:  Colley seems to be the only pollster that presents all the formula of his ranking methodology, so you may duplicate his system, and check his validity. The other pollsters leave some of their details left to this writer's imagination. All pollsters do give some literate input as to some of the specific that makeup the computer mix, like team record, location, strength of schedule, etc., but it seems that each of these items do have the possibility that each of the pollsters are looking at these categories slightly different, or at least with variables in their subsets.

Yes, they do have biases, but remember, their biases don't change. They are set for the entire season. Ther real question is, How can a team set their particular game plan to satisfy ALL of the preset biases of all the Computer Pollsters?

 
 
 
 
The views expressed in this blog are solely those of the author and do not reflect the views of CBS Sports or CBSSports.com