Seeing Luck Turn Around

One stat I like to use for seeing how teams stack up with what they "should have done" is Pythagorean expectation. It's one of the few stats with predictive value that Phil Steele doesn't include in his magazine. He does look at extreme turnover margins, net close wins, and extreme yards per play rates, so this one would fit right in.

Without getting into the math too much (click the link above to see it), it basically takes points scored and points allowed and computes a winning percentage that you'd expect for an average team with that differential. Multiply it by total games played and you see how many wins they should have had on the season. The difference between that number and the total actual wins is referred to as luck, as it is one way to measure the good or bad fortune that teams run into.

That might sound imprecise or like some kind of math nerd voodoo, but as I said, it does have some predictive value at the extremes.

Since 1998, teams with luck of -2 or less, which indicates that they won at least two fewer games than they probably should have, win more games the following year right at about 66% of the time. They win more games or stay the same at a 76.6% clip.

Five teams fit that profile in 2011, with the worst luck coming first: UCF, USF, Temple, Vanderbilt, and Army. It's a good bet that at least three of them will increase their win total this fall. USF (luck: -2.76) and UCF (luck: -3.59) will probably be two of them because all seven previous teams with luck of -2.75 or worse improved. Army is another strong contender because it only had three wins last year, and nine of the 11 hard luck teams with three or fewer wins in the span improved the next fall (with the other two maintaining the same win count).

On the other side of the coin, teams from 1998 on with good luck of at least 2 see their record decline the following year 78% of the time. They win fewer games or stay the same at a 87.8% rate. The relationship here is stronger, I suspect, because sometimes really bad teams can get really bad luck scores by being really bad. It's harder to register really good luck without actually having some good luck.

Anyway, the six teams from last year with luck of at least 2 were, in descending order, Kansas State, Marshall, Auburn, Ball State, Northern Illinois, and Baylor. Ten of the 12 teams with luck of at least 2.50 won fewer games the next fall (one improved by one win, the other stayed the same), and that net catches K-State and Marshall. Of the 25 teams with at least eight real wins and luck of 2 or more, 23 won fewer games the next season (the other two had the same win total). That captures all of these teams except Ball State.

So to bring it home, Vanderbilt probably won't see its win total decline over last year's and Auburn probably won't see its win total rise over last year's thanks to luck evening out over time. There's no rule that it has to even out in just a year's time, as some bad luck teams get worse and some good luck teams get better, but the trends are fairly clear here. I wouldn't bet against them.