Basic Statistics: Army Football & P-Wins

Before we talk football, I should note that both the Army Men’s and Women’s Swim Teams opened their seasons 1-0 last week with wins against UMass.  Army has an outstanding plebe breaststroker named Ty Dang who put up a mind-boggling 1:58.7 in his first dual meet, just a second off of John VanSant’s iconic NCAA-winning Academy record of 1:57.65 from way back in 1984.

Swimming has gotten amazingly fast over the past twenty years, to the point where I hardly know what to think of it anymore.  Nevertheless, Army has two very good swim teams, and that makes the future an exciting time.

Basic statistics at 4-2.  Click to enlarge

Following their win against Lafayette, Army Football now sits at 4-2.  ESPN has the Black Knights standing 70th in their Football Power Index (FPI), while Pythagorean Wins (P-Wins) gives them at an outstanding 87.4% winning expectation, good for approximately 10 wins, all things being equal.  Of course, all things are never equal, hence the purpose of this point: to explain a bit about what these “simple” statistics mean.

P-Wins formula.  Click to enlarge.

Baseball writer Bill James invented the idea of P-Wins as a way of determining how much a baseball team’s season is influenced by luck.  He said, essentially, that a team is as good as its winning expectation, but that all teams are influenced by their play in each specific game.  P-Wins therefore tells us whether a team is over- or under-performing.

P-Wins a useful concept, but it comes with some important caveats.  The first is obvious: it works a lot better with baseball because baseball has a longer season, yielding a much larger sample size.  Small sample size is the bane of statisticians everywhere, and I’d argue that with six games in the books, we’re just barely to the point where these statistics are actually useful.  The other caveat is that P-Wins assumes that either every point is scored equally or that our sample size is large enough to balance the “easy” points with the “tough” ones.  P-Wins therefore works better with professional sports leagues because the quality of play is much more even in the pros.  For example, P-Wins is very useful when analyzing the NFL.  I would argue that it’s probably the best basis we have for playoff betting and/or end-of-season power rankings.  

In college football, however, using P-Wins can get a little fraught.  We wouldn’t, for example, compare Alabama’s P-Wins score to Army’s.  Their differences in strength of schedule are too large to make the comparison effective.  This is why I’ve included ESPN’s FPI rankings as well.  Taking P-Wins and FPI gives us a better idea of how these teams stack up.  Add in yards-per-play, both offensive and defensive, and we develop the basis for effective comparison head-to-head.

Army has won big four times and lost twice in close games.  P-Wins suggests that despite their 4-2 record, the Black Knights are actually under-performing.  This is especially true when we consider that ESPN currently has the over/under set at ~7.5.  Based solely on Power Rankings, we would expect Army to beat North Texas, Air Force, and Morgan State, with the Navy game looking very much like a toss-up.  P-Wins suggests that the Black Knight’s ceiling is considerably higher, but importantly, these are just gross measurements.  Also: Army is a young team, and not particularly consistent.

Moreover, teams stack up according to specific matchups.  As an example, P-Wins and FPI both suggest that Army should win big against North Texas, but a look at their various strengths shows that the Mean Green have a hot young quarterback and a very good emerging passing game, which helps quite a bit against a team who specializes in defending the run.  

There’s not much to say to that; it’s why they play the games.

Go Army!  Beat North Texas!!!