Computer Ranking Methodology


In short: The AMSTS Computer Rankings are an advanced iterative loosely-ELO based system that weights wins and points to definitively rank teams, where a unanimous first place team possesses a ranking of 1.

 

The longer, less technical version: The AMSTS Computer Rankings take a number of variables, primarily centered around the result of a game, to create a profile of how “good” a team is, relative to the teams it has played.  As it does this, it calculates the margin of error between the team’s ranking and the expected result. It continues to recalculate, in iterations, until the error becomes zero. It does this independently within the “win” ranking and the “points” ranking, which comprise of the two components that make up the final computer ranking. The win ranking is weighted more than the points ranking on the principle that “good teams figure out ways to win ballgames” but the points record serves as a check to ensure that teams that decisively win games are rewarded for their good performance.  Unlike an “ELO” system, when you play your opponent is irrelevant.  Beating the eventual #1 team in Week 1 is just as valuable as beating them in the National Championship Game.

 

Phrased another way: The wins ranking loves teams that win against teams that win (against teams that win, all the way down the line.) The points method does the same thing, but with blowouts (in general). They’re then aggregated, and if a team is #1 in both win and points ranking, they will be ranked 1.0 in the total ranking. Teams that are ranked 1 in one poll but not the other will typically have a ranking in the neighborhood of ~.95, depending upon how close they are to being ranked 1 in the other poll.  The worst teams tend to end up around -1, historically bad teams (see: Grambling 2013-14 Basketball team) can touch -1.25 in overall ranking points.  Of course, a team at 0 is “average” but this tends to be around the 55-60th percentile due to out of Division games (in both college football and basketball).

 

Frequently asked questions about the rankings:

Q: How come MY TEAM is ranked lower than MY RIVAL TEAM even though MY TEAM beat MY RIVAL TEAM three weeks ago?

A: The ranking system factors in the entire body of work of a team, as well as the team’s opponent’s body of work, and the team’s opponent’s opponent’s body of work, on and on. The suggestion that a team cannot lose to a team that is worse than themselves is flawed, and using the 2015 NCAA Football season as an example, let’s take a look at using that logic to justify Otterbein University, a Division 3 school, being the best college football team in the nation:

Fact of the matter is, the ranking thinks that your rival team is better than yours simply because of how they have performed on the field and/or court.

Q: You’re clearly a fan of INSERT TEAM HERE, because they’re completely overranked!

A: In the interest of full disclosure, I will tell you that I am a Purdue Boilermaker. However, the rankings have no influences that would strengthen Purdue, the Big Ten, or harm any rival schools that you may feel I would try to harm. The computer rankings are a program, and their goal is to be free from judgment, as it helps me understand the sports world a little better as well, whereas having a ranking system that simply reaffirmed my beliefs would be silly.  And in case you don’t believe me, take a look at Purdue football’s rankings from 2013-2015.  They were, uh, not treated with kid gloves, to say the least.