Getting the right statistics

After a game, the visceral memory often fades leaving only statistics, which may end up being misleading.

I’ve seen a lot of content on the internet and elsewhere suggesting that people are looking at the wrong statistics from Stanford’s game with Duke. They all make similar errors.

1) Here’s a piece in the Stanford Daily claiming that Stanford’s secondary encountered “pass-defense woes.”

2) Jon Wilner writes, “… Nunes’ completion percentage after two games is just 57.1. It needs to in the low-to-mid 60s.”

3) In an otherwise good review and breakdown of Stanford’s prospects versus USC, Todd Husak writes: “Stanford bounced back from a shaky week 1 performance to blow the doors off of an overmatched Duke team, although the stats don’t necessarily show that. The Blue Devils outgained the Cardinal, possessed the ball for eight more minutes, threw for 363 yards, and ran 26 more plays (but lost 50-13). However, Stanford dominated field position thanks to four Duke turnovers and a 76 yard punt return by Drew Terrell which led to one fewer possession for Stanford’s offense. Also, nearly 200 yards of Duke’s production came in the 4th quarter, when the game was already out of hand and Stanford turned to the reserves on defense.”

Husak ends up getting it right, though he doesn’t explain the reason why. The reason, here, is that total yardage isn’t really relevant. A team can accumulate more yardage from another team by having worse field position, or having more possessions to accumulate them. What matters, ultimately, in a football game is efficiency rather than production. Generally football teams have equal numbers of possessions. Probably the best way of assessing overall how an offense or defense is doing, then, is yards per play.

And that stat reflects perfectly well what happened on Saturday: Stanford domination. Stanford’s offense averaged 6.3 yards per play; its defense allowed 4.3 yards per play. The offense was about a half-yard below its average last year; not great, but certainly a good performance. Such an average, extrapolated over a season, would’ve placed Stanford 4^th in the Pac-12 last year, above Washington and below USC. The yards per play difference would’ve placed Stanford in even better stead this year than last. In other words, a dominant performance without considering special teams play, which if not special was pretty darned good.

The same analysis shows why Stanford had nothing resembling pass-defense woes, garbage time included or not—with garbage time stats included, Duke only averaged 5.7 yards per pass. Renfree, the starter, averaged 5 yards per attempt. Both these statistics are awful. These statistics only reflect woes if you expect your defense to be a mutant combination of the ’87 Bears and actual bears.

As to Wilner’s criticism, that Nunes’ pass completion needs to be (we’ll say) between 60 and 65 percent or so, the standard is completely arbitrary. Having such a completion percentage is helpful for some things, but it depends on how you view it in context. For example, Sean Renfree’s completion percentage against Stanford was 70 percent. And yet it’s hard to come to a reasonable description of the game that doesn’t involve Stanford’s defense dominating Duke. The reason, of course, was that Renfree’s passes were not exactly threatening.

Nunes, on the other hand, passed for a stellar 9.2 yards per attempt. That it came accompanied with a so-so completion percentage reflects the amount of attacking Nunes did downfield—exactly the thing everyone claimed they wanted to see from the offense. That Nunes was able to achieve such production means he was pretty darn successful in doing that. Of course, an offense based on that will be prone to booms and busts absent a low-variance, consistent option.

A perfect example for how an offense can succeed in such a situation is Andrew Luck’s redshirt freshman year. Luck only completed 56.3% of his passes, which would seem to be in a danger zone. Yet Luck averaged 8.9 yards per attempt, the class of the then-Pac-10. Luck’s completion percentage, rather than signifying failure, actually signified success. The key was that the offense had low-variance, consistent options to which it could turn to get 4, 5, 6 yards when it needed it. That was known as “Toby Gerhart,” and to a lesser extent, “Andrew Luck’s legs.” (Luck ran for 358 yards that year.) The question is whether the 2012 version of the offense—if the plan is to have Nunes slinging the ball downfield as against Duke—has such a consistent option. Given the laudatory comments about Stepfan Taylor in the offseason , the guess is the coaches think they do (and, obviously, his name is Stepfan Taylor). We will see if this is accurate. Against Duke, the running game was not consistent or productive enough to serve as that option (and Nunes does not have the same running proficiency as Luck did.)

So, yes, the stats and your eyes actually agree: a dominant performance for Stanford, though one not without flaws. If the team’s line can improve as it did in 2009 and 11, the team will be in excellent shape by the end of the year.