In the NFL, it’s a numbers game
Even before analytics became a controversial obsession, statistics measured performance
There was a controversy about analytics last week, which is to say that 2017 is so far a lot like 2016. And 2015 and 2014, and I’ll stop now.
This time it was the ESPN host Mike Wilbon, a noted analytics skeptic, who took the occasion of the NFL MVP discussion to launch a Twitter broadside at statistical analysis. Or, “numbers obsessed chumps,” as he put it rather less delicately. His argument was that New England’s Tom Brady was a better MVP candidate than Atlanta’s Matt Ryan, despite Ryan having the better statistics. Wilbon said the chumps need to “assess IMPACT on the game, which isn’t explained solely by numbers.” You can tell he’s serious because he dropped an ALL CAPS in there.
But here’s the weird thing about this particular edition of scouts versus spreadsheets or numbers versus the eye test or however you want to frame it. It doesn’t even make sense. Matt Ryan had an unbelievable statistical year: 38 touchdowns, seven interceptions, 117.1 passer rating. So did Tom Brady: 28 touchdowns, two (!!!) interceptions, 112.2 passer rating.
Even using an advanced metric like Total QBR, it’s a wash: 83.4 for Ryan to 83.1 for Brady. There are arguments to be had about the two, such as whether Brady’s four missed games are too much for an MVP, or whether Ryan is too much the product of an explosive offence that allows him to pile up numbers.
Was Brady the beneficiary of a butter-soft schedule? (Yes). Would Ryan have been so good if he played on, say, Green Bay? (No). But none of this is about the modern use of analytics that gets so many so angry. How else is one to determine an end-of-season award like the MVP without considering the accomplishments of the candidates on the field? Should Brady and Ryan be measured on the basis of their haircuts and chin dimples?
Wilbon’s blast, though, is of a piece with a lot of sentiment in sports analysis: if you disagree with an argument, it must be because the argument is based on those fancy-pants analytics with silly acronyms. It’s like “analytics” has come to mean “any position with which I do not agree.”
Sportsnet’s Elliotte Friedman told a story of a conversation he had with an NHL general manager about PDO, a stat with, yes, a funny acronym that explains a team’s shooting luck. This team, Friedman explained, had a low PDO, which suggested that over time, their fortunes would improve because shooting luck tends to balance out, but the GM became more leery as he pondered how he would explain statistical regression to his bosses: “What the (expletive) am I supposed to do with that?”
PDO has also come in for some serious eye-rolling as Columbus rolled through its recent 16-game win streak. The Blue Jackets have an excessively high PDO — second in the NHL — but there was no lack of analysis in recent weeks that suggested it was unfair to ascribe too much of the Columbus streak to luck.
Teams make their own luck, and all that. But there is nothing new about the concept of good shooting fortune. Coaches have been saying things like “we deserved a better result than we got tonight” since the days when they wore jaunty fedoras. But bring up PDO, and you risk a wave-of-the-hand-dismissal as it being a silly new stat.
Kyle Dubas, an assistant GM for the Maple Leafs, once talked about trying to explain some analytics concepts to his bosses’ bosses at MLSE. He knew that the term “PDO” — originally named for a hockey blogger — might not get buy-in, so he said it stood for “percentage-driven outcome,” which is meaningless but probably sounds like familiar business jargon to someone who sits on a corporate board. That story has always struck me as telling because it explains how the resistance to analytics — to numbers, that is — is often like a language barrier.
Bar Patron: “The Blue Jackets are unstoppable.”
Bar Patron: “Well, a little lucky, too. Their PDO is.”
Bar Patron: “I don’t speak Russian. Don’t waste your time.”
But new statistics aren’t inherently at odds with old statistics. They are just more of the same thing. More data, more information, with which to evaluate teams and players. In some cases, new research has proven certain statistics to be better indicators of future success than traditionally believed, but that’s not a revolutionary point. Or at least, it shouldn’t be. If you are trying to consider how an athlete has performed, having more numbers is more useful than having fewer numbers. It is weird that this is still a thing.
In 1974, the NFL MVP was Ken Stabler, who led the league in touchdown passes. In 1975, it was Fran Tarkenton, who was tied for the league in TD passes and led it in completion percentage. In 1976, the MVP was Bert Jones, who led the league in passing yardage and had more than twice as many touchdown passes as interceptions.
Why, it’s as though the voters back then paid attention to the numbers.