In Defense of Analytics

Introducing our newest contributor to the KCSN Substack, Joseph Hefner

“All models are wrong, but some are useful.”

This aphorism regarding statistical models has been around since at least the 1980’s and remains just as true today as it did then. Analysts, statisticians, scientists, all of them are trying to create a model that perfectly represents reality. The dirty truth is this: statistical models will always fall short of the complexities of reality. No model can account for every possible thing that reality can throw at you. This is true in science, in economics, in the stock market, and yes, in the NFL. All models fall short, but that doesn’t mean that they’re not useful.

When we try to figure out how good a team is, when we look to compare our team against the rest of the NFL, there are a host of stats to choose from. We can look at records. We can look at total yards for an offense, and total yards given up for a defense. We can do the same with points scored and points given up. Each of these metrics tells a part of the story, but none of them tell the whole story.

If team record told the whole story, we’d be forced to believe that the Chiefs and the Titans are on par with each other, and a game at a neutral field between them would be a true coin flip. Instead, we see Vegas lines putting the Chiefs as 10.5 point favorites. It’s a home game, but that only counts for about 1.5 of the spread. Total yards from scrimmage tells us that the Lions have the 4th best offense in the NFL. Head to head games tells us that the Colts are better than the Chiefs. 

The moment you stop and say, “Hey, wait a minute! There’s some context you’re missing here!” In that moment, you become an analyst. In that moment, you begin to apply analytics to the discussion. You start to say that the Chiefs were missing a kicker, that Skyy Moore muffed a punt, that Chris Jones got a flag that flies once every thousand words spoken on a football field. You too are trying to look through the trees to see the forest.

It’s the blindfolded men touching the elephant. One feels the trunk and thinks it’s a snake. One feels it’s leg and thinks it’s a tree. One feels its tail and thinks it’s a rope. But when they take off their blindfolds, they all realize their mistake. They were touching only one part, and thinking that explained the whole. Taking a step back, and taking off the blindfold allowed them to see the whole context.

That’s all that analytics is. It is just an attempt at adding additional context to the discussion. It is just saying lets take your “snake”, and put it together with my “tree”, and see if that helps us make more sense of the picture. Analytics is context. It’s information. It’s data, applied to a specific situation, in an attempt to gain a better understanding of it.

All of these “advanced stats” (QBR, Passer Rating, EPA, CPOE, DVOA, etc.) are just attempts at combining different parts of the elephant into something that tells us more than just one piece does. None of them manage to capture the full essence of the elephant, but good advanced stats can get us closer to that goal. Good advanced stats are still “wrong”, in that they fail to give us a true picture of the NFL at any given point, but they are still useful in pointing the direction toward that truth. And a good advanced stat does a better job of that than simple win-loss records, or total offensive yards, or other simple counting stats. 

My name is Joseph Hefner. I am the newest contributor at KC Sports Network. I am going to be writing a series of articles on analytics and advanced stats. What they are, how they’re used (and misused), why I think they are often presented as the “boogieman” of the NFL. You’ve probably read some stuff like this before. Maybe you’ve interacted with some of the analytics crowd on twitter (I’m part of that crowd!!). Maybe you haven’t, and are wondering what NFL analytics is, and how it’s used. Hopefully this series can help bridge some of the gap between the “NFL Analytics” crowd, and the “NFL Football” crowd. I know that the “NFL analytics” crowd can get a bad rap. I’m hoping it won’t be that way here, on this Substack, in this series.

So what I will do is present this information to you in an (always) respectful, (hopefully) engaging, and (probably) informative way. What I won’t do is ever speak down to you from my white analytical tower. Can I be honest with you? I don’t care how you enjoy your football. I don’t care if what gets you going on Sundays is a run play on 2nd and 9 that goes for 12. I don’t care if what you love is to see the defense just absolutely wreck an offense and we win 12-9. Just enjoy the game! It’s not my preference, but you’re not me!

I love analytics. I love the NFL’s big data bowls they’ve been doing. I love the analytical work that PFF puts out (the articles, not grades). I think it’s incredible information. You don’t? That’s fine! Tell me about it in the comments!

Check out our new collection with Charlie Hustle by clicking the picture above!