God girl grill gridiron

Thursday, October 8, 2009

Method for Quantifying Whole-Team Performance

The question is, what’s the best way to look at a team’s performance on the whole? We basically have three important units by which we measure a football squad. The first – and most ultimately important – is the scoreboard. The second is yards, either yards given up by the defense or yards claimed by the offense. And the third is turnovers. Being an engineer and seeing three units, my impulse is to combine the three into some useful metric. According to this article at Sports Quant, fumbles and interceptions both result in about the same amount of lost points or yards. Smart Football reader Brad Eccles makes the case that fifty yards is a statistically-reasonable penalty for either a fumble or interception – that is, if you fumble, subtract fifty from your offensive yardage total. I’ve also seen (though I can not remember where, so let me know if you run across that number or a better one) that fifty yards is an appropriate bonus for plays that result in a touchdown. This seems about right to me. Thus, we can convert turnovers and touchdowns into yards and express a team’s production using a single unit. Moreover, what’s good for me is bad for my opponent and vice-versa: penalties against Auburn should be recorded as negative yards, and penalties against opponents as positive yards.

This includes the offense and the defense: the offense’s forward progress can be expressed as positive yardage, and the defense’s as negative yardage. As we learned last year, the offense and defense are not distinct.

I’ve borrowed another idea from Smart Football. There, he describes a method of evaluating a play’s success called the Sharpe Ratio. Now, I have my own issues with the Sharpe ratio – dividing by standard deviation is not a good idea when, in particular, most running backs’ success depends on their big-play potential. However, one important thing contained in that statistic is the evaluation of a play in comparison to its risk. We could theoretically run a short screen, a slant, or a quarterback draw every single play and gain a dependable, undefendable two yards every single time. While this would not be an effective strategy, all plays must be evaluated against this standard. For instance, an incomplete pass is essentially two yards given up forever. Thus, the yardage that results from each play is adjusted down by two.

I usually make some judgment about what constitutes “garbage time” and ignore that yardage. If one team is clearly bleeding clock, they’ve essentially stopped running their offense and it shouldn’t count. Also, I ignore what I consider to be uncoachable plays or dumb luck, like interceptions returned for touchdowns – these don’t count as an interception and a touchdown, just as one or the other. If the defense gets the ball, they’ve won and a win is worth fifty. Same thing for special teams touchdowns.

By this method, we can chart out our team’s progress over the entire game, in both phases of the game, as a single function. This is just what I’ve done. I copy the play-by-play at (more reliable than Rivals) into a text file, import that text file into a spreadsheet, and use the wonders of MS Excel to semi-automatically create a running list of the plays that are run, the players who run them, the down, the distance, the location and the result. Then, once I have an adjusted yardage for each play, I graph a running sum of offensive and defensive yards – this is essentially a running, risk-adjusted yardage differential. It gives some idea of the arc and flow of the game, and unites the offense’s and defense’s statistics. The goal, obviously, is to be in the positive by the last whistle. Granted, this does not reliably correlate with a team winning the game – field goals, defensive touchdowns, and special teams touchdowns are not included. But that’s not the point – I just want to see which team is most effective when they get down to work.

(I originally posted this at The War Eagle Reader, where you can see the results.)

No comments: