NFL Free Agency 2011

# NFL Statistics: Why Advanced Math Doesn't Apply To Our Perception of Football

J. Meric/Getty Images
Caleb GarlingCorrespondent IApril 6, 2011

I love football, and I loved the years I played. I also love nerding out (my degree is in molecular biology), and I love math, statistics and a good, tidy summary of the world around us just as much.

And yet, I have followed the NFL since the birth of neon and I still have no idea how QB rating is calculated.

I know it involves yards per pass attempt, touchdowns, interceptions, some arbitrary range for a couple other metrics and that’s about it. Not that the math seems hard, I’ve just known that if a quarterback isn’t consistently putting up a rating above 85, I probably won’t be sad to see him go.

And quarterback rating is a caveman compared to the growing number of advanced metrics. Over the last decade, football’s aficionados and statisticians have looked to baseball and seen all the innovative metrics Bill James and company have developed. These metrics are cool; but more crucially, they mean something. "Percentage of swings out of the strike zone" and "ground-ball rate" make intuitive sense because they apply to aspects of the game we can see with our eyes.

But what we’ve really learned (thank you, fantasy sports) is that baseball is (almost) like nine guys playing golf at the same time; each player’s actions exist in a vacuum that we can (almost) distill. Over nine innings, a right fielder may get four at bats, field two line drives, put away three fly balls and throw out a runner tagging at second. The rest of the time he could shoot the bull in the dugout and pick daisies. (Right, Manny Ramirez?)

There were exactly 10 instances where we can examine his performance—in isolation. (Almost.)

So, analogize with me...

Imagine dropping one rubber ball in a plastic bin. We’re going to observe how it careens and eventually bounces to a stop. Standing with us is a guy with a PhD in mathematics and physics. If we asked him, could he accurately measure and predicate how that ball is going to bounce, roll around and come to a stop if given the right tools?

Sure. It would take a lot of work, but he could do it.

Now think about 11 red balls and 11 white balls bouncing and colliding in a bin. Could he do it now?

Actually, yes. The number of variables explodes exponentially, but all the actions and interactions of those balls are still measurable and predictable. It might take Watson’s assistance, but it could be done.

The difference is that we, as mere humans, can’t conceptualize the second scenario. It’s too much for our brains. One ball? Sure. It goes here, bounces to the right, hits the ground, hits another wall, then comes to a stop. I get it.

22 balls at once? It’s just blurry.

And that’s the difference.

In football, statistics like Defense Adjusted Value Over Average, Adjusted Line Yards and even (comparatively simple) Quarterback Rating measure the interactions of those 22 red and white balls. They may measure an outcome accurately, but would we ever care?  (If we don’t have a gambling addiction.) More importantly, can we ever care?

Fans can’t talk about them. When debating sports, have you ever used a football statistic that required calculation beyond a simple average or percentage? Would you ever base an argument off quarterback rating? No. You’d sound like someone who’d just learned about football on Wikipedia. The number is a black box.

Even if we all took the time to learn the math, it’s not like we’re going to counter-argue a buddy’s prediction because we think one of the coefficients needs tweaking.

Yet, for baseball, you’d sound pretty damn smart if you stuck up for a struggling pitcher because, looking at the numbers, his fly ball ratio is down while his ground-ball and called strike ratios are actually up; he’s pitching better than his recent tough outings would suggest. In this case, we’re still discussing actual situations in the game. The math augments what we observe.

The advanced statistics in football don’t have the same digestibility.

You may tell me that Tom Brady was actually not as good at quarterback last season as the standard statistics suggest because of his Adjusted Net Yards per Pass Attempt, and I could say:

“Okay, good sir! Let me counter your argument by running my algorithm of how Brady interacted over 16 games, each with 11-plus players in hundreds of combinations, running a playbooks of different formations, against over a thousand combinations of defenses and personnel, and....zzzzzzzzz. Sorry, I blacked out.”

I don't discount that front offices and assistant coaches need these statistics to pour over in preseason and predraft meetings, but as fans, wanting to argue about the game, while feeling like we're still talking about the game, it's too many bouncing balls.