Cowboys vs. Lions: Detroit’s Trends
At ABC, I posted some trends on the Lions:
Numbers never lie.
No really, despite what you might hear, they don’t. Math, as a flawless abstract concept, is pretty cool like that. The numbers, by their very nature, are perfect.
But people lie all the time. And so I can twist and frame numbers in pretty much any way that I want to get my point across. If I were born 20 years earlier and had an unusual fixation with “establishing the run,” I might point out teams are (insert awesome record here) when they run the ball (insert high number here) times, so you need to run the ball.
The numbers didn’t lie and they aren’t “wrong” in any sense, but I manipulated them in such a way that I could back my preconceived notions. I established a correlation but proposed a causal relationship that doesn’t exist. There’s nothing wrong with the math, just my interpretation of it.
Sorry, but Stuart Scott was wrong. Don’t hate the game, hate the player.
But I’m not in the business of using numbers to deceive people. Instead of framing math around my arguments, I want my opinions to be a reflection of the math. I want to properly interpret all of the data that’s so readily available these days to provide a deeper understanding of the Cowboys and the NFL.
One way that’s possible is by identifying predictive ability. The fake “establish the run” stat I gave you isn’t very useful—it’s misleading, in fact—because it’s not predictive of what wins. That’s why we see teams that pass the ball a lot early generally have a lot more success than the running teams (with the overall run-pass balance evening out as teams that are already winning run the ball and teams that are already losing pass it).
By establishing a stat’s predictive ability, we’re basically measuring how much it “lies.” That allows us to sort through the noise, like time-of-possession (also a result of winning), in favor of the signal.
Detroit Lions By the Numbers
I hear the Cowboys have a game against the Lions this week, so I guess I should stop ranting on the philosophy of math and take a look at that, huh? Okay, fine.
51.9: Lions’ first down run rate
You might have noticed that I tend to analyze first down stats quite a bit and third down stats very little. The reason is that, for the most part, first down is standardized. Whereas third down plays can be of any distance, around 95 percent of first down plays are the same: first-and-10, typically near the middle of the field.
As a quick side note (I promise I’ll talk about the game soon), that’s one reason that third down conversion rates are useless stats. Yes, you want a high conversion rate. Yes, third downs are really important in each game. But the goal for many teams is putting themselves in “manageable third downs” when it should be maximizing offensive efficiency.
Running the ball on first and second down to set up a bunch of third-and-three situations might lead to a high third down conversion rate, but it’s not beneficial to the offense. You know what’s better than converting a high percentage of third downs? Not facing third down at all because you didn’t run the ball on first and second down.