Birds' Nest
A home for everything Eagles

On Misleading Sacks, Stats

Email
|
Print

    NEWSLETTERS

    I'm a big fan of the Moving the Chains blog by Sheil Kapadia. It's nice to see a member of the Philly sports media embrace statistics and the fan's perspective.

    Which is why I'm conflicted about Sheil's post earlier this week, "Why Vick was sacked 34 times." On one hand it shows some great initiative and effort to go back through all the of the Eagles sacks in 2010. But the analysis lacks context, and therefore the numbers are misleading.

    Kapadia broke down Vick's sacks by the number of opposing rushers. He was sacked 12 times with four rushers, 11 times with five rushers, and seven times with six or more rushers. That's interesting information, but it's incomplete. Without any context, it looks like the Eagles gave Vick the least protection when faced with minimal pass rush. But that's not true.

    I don't have the information for last season yet, but I compiled Football Outsiders information on number of rushers from 2009 (in FOA 2010). Teams averaged four man rushes 60 percent of the time. They blitzed one extra player on 24 percent of pass plays, and rushed two or more (six players plus) just under 10 percent of the time. In other words (not to blow your mind), most teams don't blitz anywhere close to the majority of the time.

    It's important to note such a fact when you're listing total sacks. The Eagles likely faced four man rushes about 60 percent of the time — but only gave up 40 percent of their sacks on those plays. Meanwhile, about 60 percent of Vick's sacks came on blitzes, even though blitzing was likely no more than 35 percent of the time. Suddenly, it looks like blitzes were a much bigger culprit for those sacks than we originally thought.

    Now perhaps the Eagles were blitzed more often than most teams, or there was some other anomaly. The truth is that just looking at single set of plays that resulted in sacks is a crude and selection bias-inducing measurement. It might look like there's correlation there, but you have to step back and consider all of the data for a more complete picture.

    I'm hopeful that Sheil's follow up post will give more context along these lines.