About BracketBot


The genesis of BracketBot actually began with football. The first ranking system developed by Zach using this technique was applied to the 2013 NFL season! College football came later, but only that winter did BracketBot start to blossom during basketball season.

Recursive calculation

Ratings are continuously recalculated whenever new data is available. Whenever new games are completed, BracketBot uses the results of those games to try to reconcile them with the results of every other previous game this season.

Two different rating systems

The final rankings that you see on this site come from a combination of two different philosophies: a backward-looking assessment of how good a team's résumé is, and a forward-looking assessment of how likely they are to win in the future. Currently, the proportion is three-quarters résumé, and one-quarter left to the offensive and defensive ratings.

Additional tools

BracketBot is equipped with a number of different tools on top of ranking teams: it can put together a bracket, acting as its own Selection Committee, correctly identify the winners of tiebreakers for conference seeding, assess schedule strength using my own metrics as well as RPI, and simulate the results of future games.

Is this based on preseason data?

Nope! Although I do start with preseason rankings, compiled and combined from a number of different sources, that's only so the first month of the season doesn't look like total nonsense. By mid-December, all preseason influence has been removed from this system.

What do the numbers actually mean?

An average offensive or defensive team receives a rating of 1000, and the scale is multiplicative: a team with a rating of 1200 is 20% better than the average team, and a team with a rating of 700 is 30% worse than the average team.