Plus-minus is sort of the Holy Grail of individual basketball statistics. That's because it captures everything that is happening on the court, whether that's making baskets, setting a good screen, or undercutting shooters (I'm looking at you, Wisconsin). The problem, of course, is that capturing everything happening on the court is quite noisy. In order for a plus-minus statistic to have value, there needs to be a lot of data and the analysis of that data has to account for all 10 players on the court.
In the NBA, this analysis is done using regression, because there are only 30 teams, and those teams play each other quite often over an 82 game season. This makes it possible to compute what is called adjusted plus-minus; that is, plus-minus for an individual while accounting for all other players on the floor. Dan Rosenbaum at 82games.com explains the nitty-gritty here.
However, even in the NBA the data is noisy. To help correct for some of this, Rosenbaum also estimates what the player's adjusted plus-minus would be, given their regular box score statistics, and calls it adjusted statistical plus-minus. This is then blended with the actual adjusted plus-minus to yield Rosenbaum's final figures.
The problem with this approach in the NCAA is that there simply isn't enough data to yield decent results for regular adjusted plus-minus. There are too many teams, playing too few games against too few opponents. Ken Pomeroy's treatise on the topic is well-written and explains some issues.
Despite this, statistical plus-minus should still give us a pretty good estimate of what a player's plus-minus should be given his regular box score numbers, with a few drawbacks. First, we have to apply the NBA formula to college, because there is no good way to obtain an NCAA-specific estimation. Second, the usual caveats about minutes played and sample size still apply.
Ultimately, I believe adjusted statistical plus-minus will be most useful in estimating a player's defensive contributions, and perhaps offensive contributions for players who use very few possessions. This is because KenPom gives us a very fine-grained look at offensive contribution, but relatively few defensive stats.
In my next post, I will be looking at the ASPM for Michigan State players in 2013-14. Please leave questions and/or comments below.