Voxitatis has applied a modified version of its mathematical model to predict all eight football titles in Illinois, but, like last year, we are waiting until after the Class 8A game is played to publish those predictions.
Our reason for doing this is that the overwhelming majority of hits on that page, which come from Google searches, involve gambling or betting on the games, and we don’t encourage this type of activity for high school athletics.
We’re not going to change our answers based on the actual outcomes—the computer has already printed the story. We just need to delay the publication of our predictions.
The games take place today and tomorrow at Huskie Stadium on the campus of Northern Illinois University. Classes 1A through 4A play Friday, and classes 5A through 8A play Saturday.
- Our database dump of full season records for all finalists
Our basic mathematical model
What we did in the past was to compute a statistic that would tell us about how good a team was at winning games, which comes from scoring more points than your opponents in the games. The general form of the model we used is as follows:
where σ represents a statistic that is directly proportional to a team’s overall strength. In our model, f represents the offense (average number of points scored in every game up to that point in the season), d represents the defense (average number of points given up in every game so far), and the subscripts have t for the team for which the statistic is being computed and o for that team’s opponents.
In other words, a team that gives up as many points as they make will have a σ somewhere close to 0. We adjust the average points scored by the team by subtracting the average number of points given up by that team’s opponents. What this does is something akin to giving teams that play tough opponents a higher rating, because a lower number of points will be subtracted if the team faced all tough opponents than if the team faced only very weak opponents all season.
Likewise, the number of points given up is adjusted by subtracting the average number of points the team’s opponents earned across all the games that season. This effectively means that if the team played opponents who didn’t score very many points anyway, it’s not such a big deal if this team held them to a shutout.
Over the years, we have played with the coefficients on each of the terms, but A always equals C and B and D have always been less than A and C, respectively. Two years ago, for example, we used:
- A = C = 1.00
- B = 0.50
- D = 0.25
Using those values, we predicted the winners in seven of the eight games for a “batting average” of 0.875. This year, we have put considerably more computing power behind the predictions, and the coefficients are going to be adjusted based on the results in each team’s semifinal games. That is, whichever coefficients best predict the margin of victory in the team’s semifinal game—the same values of A, B, C, and D have to be used for the team and its semifinal opponent—those are the coefficients we’ll use to get the σ statistic for the team in the final game.
Our prediction is simple: Whichever team has the higher σ will be predicted to win the final game. The greater the difference between the σ statistic for the two teams, the more likely our computer model predicts that team will win the game. This is not a “spread” or anything related to a point differential: it is a probability statistic used for comparing two teams playing against each other in a single game at a certain point of the season.