As I am sure you noticed, I posted the predicted win distributions on each of the team pages for both overall wins (based on games listed on the schedule portion) and conference wins. They give the percent chance of a team finishing with a specific number of wins based on the current ratings. Much thanks to my buddy Todd from the best analytical bowling blog in all the land (sadly now defunct) for providing the inspiration.  Ultimately, the charts may find a less conspicuous location as they are probably only for the serious mathletes out there.

I don’t know that these charts will be as interesting later in the season as they are now. As more results become known and the ratings converge on the truth, the distributions will become more narrow. Now though, it looks like anything can happen for just about every team.

I am particularly obsessed with the tails of each plot because they describe well the degree to which randomness can affect a season. (I should mention that I compute these probabilities by taking the predictions for every game and running the season 10,000 times. I do this every morning using the latest ratings so you don’t have to.) Take Boston College for example. They lost to Yale last week. Yet if the ratings are accurate right now, there’s a 3% chance BC goes 11-5 in ACC. In 10,000 runs, there was one trial where they went 0-16 and one where they went 14-2. The model isn’t saying BC is going to improve or implode in those cases. It’s the same team that is currently rated seventh-best in the ACC providing that range of results. 

If BC plays as expected, it almost surely isn’t going to win 11 or more ACC games. They’re a 50-1 shot to do that. But across all of D-I, some teams are going to pull off an improbable feat like that. By the end of the season, there are going to be teams that have misleading records – a few very good teams will have poor records and a few will have great records and not be very good.

I’m not suggesting you fully trust my model to identify those teams at the end of the year (though I think it does a good job). I am suggesting you embrace the concept that for a few teams, their record in mid-March will be more the product of randomness than it will be the product of skill.