I have a brief blurb on my sidebar highlighting how to use EM as a proxy for expected margin of a game, but I though I’d be a little bit more thorough in a blog post.
As I’ve explained, EM is a measure of expected margin of a given team against a hypothetical average team over 100 possessions on a neutral court. The model actually predicts the points/possession margin, and I multiply these numbers by 100 to get what is displayed on the site.
The EM ratings have the great feature of being a linear metric. This means that a 30 EM team is to a 20 EM team as a 10 EM team is to a 0 EM team. This feature is what allows us to use EM to predict the margin for a competition between any two teams.
The formula to predict expected margin from these numbers is as follows:
(EM(1) – EM(2)) * (expected possessions / 100) +/- HCA, where HCA is positive if Team 1 is home, negative if away, and 0 if neutral.
The average number of possessions in a college basketball game is around 70, and the average home court advantage is +/- 3.5 points, so for a simple formula you can plug these numbers in. However, to be more precise, there are team-specific numbers for these two values.
Possessions per game is dependent on the tempo of the two teams playing. A decent proxy for the expected possessions of a game is the average between the average number of possessions between the two teams. Kenpom has good (and free) numbers on this. Using the average possessions between the two teams is not the best metric, though, for two reasons: (1) A team’s average possessions per game needs to be adjusted for the tempo of their opponents throughout the season, and (2) some teams are more tempo-dominant. This means that there is a spectrum for how a team’s tempo changes per game: some teams force another opponent to play to their pace (very constant tempo), and other teams oblige to play at whatever speed their opponents play. There are some predictive models that account for this; mine is not one of them.
Home court advantage additionally varies by team, though not too significantly. Distance explains the majority of the HCA value. As expected, Hawaii has the strongest home court advantage at + 4.5 points. Other midwest teams and Pacific-coast teams have strong home court advantages. Credit to Ken Pomeroy again for these numbers, which I actually use in my ratings. Home court advantage varies between 2.5 and 4.5 at the extremes, so if the game of interest is particularly close/far, it may help to adjust your expectations by a few tenths of a point.
Here’s how the rough EM formula performed for games this Saturday January 25:[table “21” not found /]