About

The purpose of this endeavour is to provide an alternative method – the Win Value Adjusted Ranking Average (WVARA) – for predicting the outcomes of college football games (and other sports where applicable), and ranking each team accordingly. There are many methods currently in use today, most notably the systems in use by the Sagarin System and the BCS computer ranking systems.

The Philosophy Behind WVARA

These systems (and others) rely on a large variety of stats to determine the outcomes of games. However, WVARA only uses one stat (or two): points scored (and allowed). Football differs from other sports in that it relies heavily on a scripted attack or defense – if the play is not well rehearsed, or if the players lack the skill to execute, this will have an effect on the entire game. Games like basketball with more fluid motion do not have this constraint (however, the case may be made that WVARA is viable for all sports). The structure of football plays allows the entire team to be evaluated as one unit. This differs from baseball and its associated sabermetric systems, where each individual player may be evaluated for his overall talent and contribution to the game. Football (and to a lesser degree basketball, et al.) is on the opposite end of the spectrum in that a single player rarely has enough talent to affect the game regardless of a lack of contribution from the rest of the athletes.

For this reason, even though WVARA works for sports like basketball, this paper will explore its conception and application as relating to NCAA D1 football.

A common public misconception is that a team is more likely to score (or allow) an average amount of points in a game regardless of the opponent. An oft-quoted pre-game statistic is average points scored/allowed per game. This is incorrect because not all teams are created equal. A good defense facing bad offenses will fare much better than it will against a very good offense, and this will skew the numbers of how many points the defense actually allows. Therefore, a large portion of the analysis will be directed at adjusting this average value to a more realistic number based on the quality of the opponent.

This system is ideally based on head-to-head matchups. Each team is given a likelihood of winning (%W) based on the opponent’s caliber. This is based on the team’s performance to that point in the season, but is not weighted based on the opponent’s performance, only on the opponent’s average scored or allowed points {as of 4/7/2013}. Because this metric is calculated for each team separately, the values may not add to exactly 100% for both participants. To aid in predicting the actual winner, an adjusted value is also calculated that takes into account both %W values.

WVARA and the associated rankings take margin of victory into account (discussed later) but ignore home team advantage. Home team advantage has indeed been shown to be a factor in the real world, but this model does not explore its effects on the data at the current time. More research and analysis would have to be done in order to factor this in.

Observed Differences with Existing Ranking Systems

Several things must be noted about the standings and power rankings. The standings are based on how well a team performed based on what was expected of that team (this is similar to the ranking systems used by 4down20.com). A very good team that beats a bad team is not awarded as many points as a team pulled an upset over a “better” opponent. Conversely, when a good team loses to a bad team, it is punished more heavily than a bad team suffering a loss. This may introduce significant discrepancies between a team’s place in the WVARA rankings and their place as perceived by more subjective polls (AP, BCS, etc.). The power ranking is based on the predictive metrics, and is a measure of what percentage of all other teams that team can beat: a simulation is run for each team against every other FBS team. This may also yield uncomfortable results, in that one team that is “worse” than another may have a higher ranking simply due to their probability of beating more teams.

The 2013.2 WVARA System

These rankings use the 2013.2 version of the WVARA formulae set. This method should be available in a longer paper detailing its development by the summer of 2014.

Statistics are taken from the Massey Ratings.