One of the most useful measurements of modern sabermetrics is Wins Above Replacement (WAR). Pioneered in baseball, the statistic attempts to calculate the total number of additional wins generated by a given player, at least compared to a “replacement level player” of ordinary talent. In general, a player who generates more than eight wins per season is of MVP quality. Five wins is All Star level, while most starters hover around the two wins mark.
Baseball, of course, is an ideal sport for the WAR approach, since the performance of its players is largely independent. Every hitter hits alone; every pitcher is on the mound by himself. In recent years, however, the WAR stat has moved beyond baseball. Bill Gerard developed a similar model for soccer players in the Premier League, while John Hollinger created “Estimated Wins Added” for the NBA. All of these statistics rely on the same basic strategy as WAR in the major leagues: they compare each player’s performance to a hypothetical “replacement,” and convert the difference into the only measurement that matters: winning.
Football, for the most part, has missed out on the WAR revolution. This is for an obvious reason: it’s really hard to disentangle individual statistics from team performance. (There’s also a shortage of individual stats in football.) Take a running back that scores on a two-yard touchdown run. Was the touchdown really due to his talent? Or was it triggered by the excellent blocking of his offensive line? And what if the play was setup by a long pass, or a fifty yard punt return? The running back gets most of the glory, but it’s unclear if he deserves it; a replacement level player might have scored just as easily.
To get around these issues, most football metrics make a problematic but necessary assumption: that all player statistics are independent of teammates. The wide-receiver doesn’t depend on his quarterback and the quarterback doesn’t need a good offensive line.* But that’s beginning to change. A new paper, by Andrew Hughes, Cory Koedel and Joshua Price in the Journal of Sports Economics, outlines a WAR statistic at the position level of the NFL, rather than at the level of individual players. While this new stat won’t tell you who the MVP is in a given year, it will tell you which positions on the field are most valuable, and thus probably deserve the most salary cap space.
The logic of positional WAR, as outlined by Hughes et al., is quite elegant. In essence, the economists look at what happens to a team in the aftermath of an injury or suspension to a starter. As they note, such events – while all too common in the brutish NFL – are also largely random and exogenous. This means that they provide an ideal means of investigating the difference between a starter and a replacement player at each position. By comparing the performance of teams before and after the injury, the economists can see which starters matter the most, and which positions are the hardest to replace.
So what did they find? On the offensive side, there are few surprises. Quarterback is, by far, the most valuable position: a starting QB that misses four games due to injury will cost his team an average of 1.3 wins. That’s followed by tight ends, fullbacks, wide-receivers and outside offensive linemen, all of whom cost their team roughly half a win for every four missed games. Interestingly, running backs appear to be rather interchangeable: when a starting back goes down, there is no impact on the team’s overall performance. Such data would not surprise Bill Belichick.
The defensive side is where things get strange. According to Hughes et al., teams are not hurt by the loss of defensive starters at any position. Put another way, the positional WAR for every defensive player is essentially zero. If a starting cornerback, safety, linebacker or defensive lineman goes down, a team should expect to win just as many games as before.
At first glance, this data makes no sense. Could a team really win just as many games with second-tier defensive players? The answer is almost certainly no. Instead, the low WAR of defensive positions is probably a testament to the power of defensive strategy. As Hughes et al. note, “defensive schemes can be adjusted to account for replacement players more easily than offensive schemes.” If a cornerback goes down, the safeties can help out, or the linebackers can drop back. The same pattern applies across the entire defense, as coaches and coordinators find ways to compensate for the loss of any single starter. Of course, if multiple starters go down, or if a defense has a lower level of overall talent, then it’s that much harder to compensate. Schemes and scheming have their limits.
That said, this data does suggest that the typical NFL team could improve their performance by spending less on defensive superstars. If you look at the average salary for the ten highest paid players at each position, it becomes clear that NFL teams treat defensive starters as far more valuable than their replacements: only quarterbacks and wide-receivers make more money than defensive ends, with linebackers, cornerbacks and defensive tackles close behind. Again: it’s not that these positions aren’t valuable or that Richard Sherman isn’t supremely talented. It’s that a smart defensive plan might be able make up for a missing defensive star. Russell Wilson, on the other hand, is probably worth every penny.
And so a statistical tool designed to reveal the most important players on the field ends up, largely by accident, revealing the unexpected importance of the coaches on the sideline. While there are numerous measurements of bad coaching – one study concluded that suboptimal 4th down decisions cost the average team roughly ¾ of a win per season – this new paper highlights the impact of good coaching, especially on the defensive side. If you have a successful system, then you might not need the best available safety or lineman. Perhaps you should save your money for the passing game.
At its best, the sabermetric revolution is not about neat answers, or the reduction of talent to a few mathematical formulas. Rather, it’s about revealing the deep complexity of human competition, all the subtleties the eye cannot see. Sometimes, these subtleties lead us to unexpected places, like discovering that you can’t effectively evaluate half the men on the field without taking their group strategy into account. The whole precedes the parts.
* Total QBR is a possible exception to this rule. Alas, its equations are proprietary and thus impossible to evaluate.
Hughes, Andrew, Cory Koedel, and Joshua A. Price. "Positional WAR in the National Football League." Journal of Sports Economics (2015)