Reading the Defense With Numbers

I played football at a reasonably high level - never got paid to do it, but spent a significant chunk of my life in and around the sport. It's been over a decade since I last played, and these days I'm more focused on family and career than keeping up with my Alma mater, let alone following the latest developments in the game. But whenever I happen to catch a game, I'm struck by how quickly I can still recognize certain patterns that reveal what's about to happen on the field. Anyone who's played beyond high school has watched hundreds to thousands more hours than they've actually played - and regardless of how thick your forehead is, some of that knowledge just sticks with you.

I hadn't really contemplated how the technical skills I've developed since my time playing could overlap with the sport until a recruiter reached out about a Data Engineering role with an NFL team in late 2024. It was a really interesting opportunity given my background. I won't name the organization, but it's one that seems to make questionable personnel choices - thus I didn't get the job. However, I met some really interesting people and it was fascinating to see what these teams are doing with the data they have access to. It was a reasonably thorough interview process, but we only touched on the surface of what I imagine is a relatively complex set of systems used to support decision making across the organization. Due to the nature of the role I was interviewing for, the bulk of the insights I gathered were focused around talent evaluation - which personally I'm skeptical of. The athlete in me still views data as supplemental to evaluating if a player just has 'that dog' in him.

In all seriousness, I was disappointed I didn't get the position, but I'm sure someone else was equally excited they did and I'm truly thankful for the experience. My curiosity had been sparked and some memories from my playing days came rushing back. Specifically from my senior year at UT.

When I'm asked why I quit playing, my typical response is "I wasn't good enough to play anymore." Which no one is going to argue with - but there's probably more to it than that. I can assure you there are plenty of instances every Sunday in the fall when a guy is on one of the 32 teams' 53-man roster that should be in street clothes and someone else bagging groceries that should be in a helmet. At the margins, they're razor thin. Sure, a guy like Aaron Donald won't be missed, but there are way more players than there are roster spots, it's just math.

The truth is, I had a nagging knee injury my senior year that couldn't be diagnosed without an arthroscopic surgery, which I put off until the end of my senior year. The day after Texas lost to (a very talented) Oregon team in the 2014 Alamo Bowl, I was in Dallas, TX undergoing a "routine scope" that turned out to be a micro-fracture surgery due to a quarter-sized chunk of cartilage that had detached from my femur below the knee cap. I didn't recover well, I was poor and didn't have the stomach to grind it out with a bum knee that somehow felt worse than before the surgery.

I don't share that to relive what could have been, but to provide some context as to the condition of my health during my senior campaign. It wasn't great. Looking back - I shouldn't have played but thanks to the following, I did:

What I believe folks like to call "Toxic Masculinity"
Toradol (IYKYK)
The fact that we weren't especially deep at my position
I understood our offense
My 3 years of experience playing against (mostly) the same defensive coordinators in the Big 12

The last 2 are the important ones.

Candidly, I still question if I was a net contributor or liability for the team my senior year. If you somehow forced me into the chair from A Clockwork Orange where they cured Alex of the ole Ultra Violence and had me watch all the tape, I don't think I'd like the answer - but it's funny what you remember naturally, probably to protect the ego. I can still recall a handful of plays vividly where I know beyond all doubt that I was the only person capable of identifying and communicating what the defense was going to do, and that had a direct positive impact.

Now, this isn't uncommon in the sport. Many will chalk it up to an intangible skill like "intuition", and when I hear stories like Troy Polamalu forcing Pat McAfee to call off a fake field goal attempt by doing something he'd never done on tape previously, sure - I think pure intuition plays some role - but that's uncommon. What's much more common is elite players studying and analyzing the opponent to limit the possibilities of what they know is going to happen on a given play based on historical observations and present context - and there are layers to that context - the deeper you go, the fewer the possibilities:

3rd and 7 to go
On the 30 yard line
2nd Quarter - 50 seconds till half
Offense down by 2, received ball to start the game
Formation
Personnel

I sometimes laugh (not out loud) at the commentary of friends and family when I watch a game. It's like watching someone talk about chess who doesn't understand how the pieces move - anything is possible in their mind. When in reality there are only ever maybe 50 plays that either side of the ball can run at any given point in a game. If you get semantic, sure there are 1,000s of variants, but running 3 down and blitzing a linebacker turns into a 4 down front the same way a 4 down front with a defensive end dropping into coverage turns into a 3 man front. It's just a different way to do the same thing.

Fifty may even be a conservative estimate. For any player worth their salt, on either side of the field at a high level, that number is closer to 10. When you get down to crucial points in the game, those moments with ample context, that number drops precipitously. Then when you have one of "those guys" (Luke Kuechly, Peyton Manning, Tom Brady, Ed Reed) they know exactly what's about to happen (there may even be other guys with the same level of foresight based on analysis that lacked the physical talent to be a "named man" - usually we call those types of people "Coaches").

The ah-ha moment for me, and how all of this relates to my career discipline, is that it's all just pattern recognition. And relatively trivial pattern recognition at that. The complicated parts are the inputs. Or at least, that's what I thought.

After I interviewed for the Data Engineering role with the "team I'm not naming," I stumbled across a dataset linked to a "Big Data" competition the NFL puts on. You're never going to guess what they call it. Yeah, it's the "Big Data Bowl."

I've actually done some analysis using this dataset, which I'd love to share, but I'm cautious. The dataset is publicly available, but we can all hear that disclaimer played before every kickoff: "Any reproduction, accounts or descriptions of the game without permission of the NFL is prohibited." The last thing I need is a lawsuit resulting from a blog post only my mother will see.

What was interesting about the data, to me, is that it's all there. Everything you would need to build a predictive model to scout opponents is available, along with a ton of feature-rich metadata about players, coaches, weather conditions, speed of movement—it's all there, for every player on every play.

When I'm talking about this with someone, they inevitably ask, "Well why don't you do it if it can be done?" My immediate answer is usually self-deprecating—but I firmly believe it already is being done by people who went to school to make rockets that shoot down other rockets, or cars that don't need drivers. Based on what I learned during the interview process, folks have been working on these types of projects for years, but we're just getting to the point where the data collection has caught up with the capabilities of what's possible with predictive analytics—and it will only get better.

The question in my mind is who will adopt this technology first and what does that look like in terms of success. Sports analytics is somewhat of a faux pau in the world of football where it's found success in other sports (Moneyball, Steph Curry, etc.). Part of that is warranted. Plenty of well intentioned smart people tried to do something smart eventually met someone like Ray Lewis - which is one of those black swan events where all correlations go to 1. Talent still wins the day. But based on my experience as a student of the game, the hurdle to finding what to look for when scouting an opponent has been lowered, massively. With the sheer amount of computing power that can be thrown at these problems, alongside the robust datasets that are being generated, it changes the game. I don't think it can make a bad team good, but it can make a team better. I'll put it this way, I don't think the team that wins the Super Bowl next year wins because they had the best predictive models to scout opposing teams with. However, I don't believe there will ever be another team that wins a Super Bowl that doesn't have a great predictive analytics platform to scout opposing team with. And I think the same is true eventually at the College level. I hope it stops there, God knows the last thing this world needs is little Timmy's dad showing him a flowchart of what coverage to expect on 3rd and short for his flag football game - but I won't be surprised.