AWS re:Invent 2025 - How AWS and Next Gen Stats Redefine Football Intelligence

🦄 Making great presentations more accessible.
This project aims to enhances multilingual accessibility and discoverability while maintaining the integrity of original content. Detailed transcriptions and keyframes preserve the nuances and technical insights that make each session compelling.

Overview

📖 AWS re:Invent 2025 - How AWS and Next Gen Stats Redefine Football Intelligence

In this video, NFL Network's Cynthia Frelund, Mike Band from Next Gen Stats, AWS's Julie Souza, and analyst Greg Olsen discuss the evolution of NFL analytics over the past decade. They explore how Next Gen Stats has progressed from basic speed metrics to advanced AI-powered insights like completion probability and coverage classification. Olsen shares how data preparation has transformed his broadcast work, with analysts providing detailed packets on team tendencies and play styles. The panel highlights the Big Data Bowl competition, which has produced over 60 NFL analytics staffers and generated new metrics like route classification and tackle probability. They discuss the balance between data-driven insights and human expertise, emphasizing that analytics serve as tools rather than replacements for coaches and players. The conversation covers real-time graphic integration challenges, the shift from 2D to 3D tracking data using Sony Hawkeye Systems, and applications in player health and safety that have reduced injuries. They stress the importance of translating complex data into accessible storytelling for fans while maintaining technical rigor.

; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.

Main Part

Introduction: The Evolution of Next Gen Stats and AWS Partnership

Should we sit in order? We sit in the order of our, so you can figure it out here. I love it. Well, I'm Cynthia Frelund. I work at the NFL Network with my friend here, Mike Band. I'm also a data scientist, and I'm really happy to be here. It's my second time here at re:Invent, which is very fun. I'd like to introduce to my left Mike Band. He's a Senior Manager of NFL Next Gen Stats Research & Analytics. You can read his title just like I can, but I would like to tell you that he's about to become a two-time dad. He's already a one-time dad. We like that. Julie Souza, she's the Principal of Sports Marketing here at AWS. She's the head of all of everything that AWS is doing in our world, at least. Julie and I also used to work together at Disney, you know, fun fact—small world. And last but not least, you know him from being the absolute best analyst on TV, and it's not even hyperbole, and I'm not saying it because he's sitting there and much larger than me, but Greg Olsen, a three-time Pro Bowler, the first tight end with three consecutive 1,000 receiving yard seasons, two-time Emmy Award winner, will be like six before he kind of hangs that one up too. Current Fox analyst, the best one, Greg Olsen too. So happy to be joined by you guys.

Good to be here. Alright guys, so it's about to be, or I guess it is going to be the tenth season of Next Gen Stats partnered with AWS, which has kind of flown by if I'm honest. We started out with all of these speeds, how fast is a player running, et cetera. Julie, what, take us back a decade ago. Tell us how this all kind of came to be. I think you're hitting an interesting point because it started sort of simply, right? The league started collecting data. They have sensors in the shoulder pads and the ball is chipped, and they're collecting things like speed, time, distance. So those were the early day analytics. But now with the power of technology and predictive tools with AI, you can start getting much more advanced in the analytics. You've seen Next Gen Stats really evolve with that, with the technology to become much more interesting, much more predictive, and much more engaging. That's where it started, where it is, and where it's going.

From Simple Speed Metrics to Complex Analytics: The Early Days

I remember I came to the NFL Network in 2016 and we would do these video carousels or these pictures and they'd say the speed that Tyreek Hill got to was 22.96 miles per hour. I mean it's fast, but like how fast? I don't even know. Is that good? How good is it? So Mike worked on all of the Next Gen Stats throughout the league. He has for a long time. His background, he can tell you, or I can tell you, he worked at the team level. He's done it with talking to other broadcasters across all of the networks, and he has to figure out how to herd all of the cats at the NFL Network as we were trying to figure out how to implement all of this stuff. So give us a little bit of what was back then and how did it kind of come to where we are today.

Yeah, about ten years ago, we were relegated to box score stats and certainly there were companies out there like Pro Football Focus, Sports Info Solutions that were doing manual hand charted data. So this meant that you could take our GSIS box score stats and come up with—GSIS is like the main service that does all of the box scores for everyone, exactly. And so you can derive stats like yards, tackles, touchdowns, the basic stuff that's been around for over 100 years. With Next Gen Stats coming along in 2015, every player, Greg was there in those early days, had two RFID chips in their shoulder pads, and that created that XY tracking data. And so about ten years ago we were just worried about storing the data, collecting the data, and we were getting into the processing and the analytics, deriving those meaningful stats from it. But it really, as you said, going from speed and distance to now much more complex data points like completion probability, what coverage they ran, it's been a really fun ride and a journey that we've had AWS be there every step of the way.

Before we ask Greg what this data meant to him as a player, which Next Gen Stat do you wish that you could have had on Greg? That we could have from his former data because I could tell you I want to know forced missed tackles. I want to know how many times you just were like see you and got passed. Not as many. You should pick a different stat. Yak there we go. I will say yes, absolutely. Just to see, you know, I'll let you know what I'm good at and I'll get back to you. Just to know the style of what Greg was doing on the field, right? Was he in the slot? Was he in line and tight? Was he out wide? How often was he targeted? How much yak did he get per play? But not just how much yak, how much yak did he gain over expectation? We've really developed a whole suite of metrics that really tell the story beyond just that basic number.

A Player's Perspective: Greg Olsen on the Data Revolution in Football

So what was data about ten years ago to you when you were at the end of your career? What did it mean to you? You know, so I came in the league in 2007. So from 2007 to 2015,

it was actually the year we went to the Super Bowl. Up until that point, we were box score stats ourselves. We would get scouting reports and evaluations of the other team. For us on offense, the coach would give the Wednesday scouting report with total yards given up, total points given up, rushing yards per game, passing yards per game, and completion percentage. I look back on it now as I prepare for games on this side of my career. I spent exactly zero minutes worrying about any of those stats, and they are completely irrelevant to determining good teams versus bad teams. There's just no context anymore.

From 2007 to 2015, that's all we were privy to. That's all we had. Some of the teams had more advanced analytics departments upstairs doing it more manually. It was a lot of game film evaluation, and they were trying to get next-level type stats, but it was all through observation. It was all manual. It was not machine learning. It was not data points. We did not yet have the geo-tracking on the field—size, speed, location, how far were you from me when I caught it, yards of separation. We didn't have any of those metrics, so you could only go based on what you see.

As a player, I retired in 2020. From 2015 to 2020 compared to where we are now in 2025 as I'm evaluating this and preparing for my games each week, we've gone from zero to 60—we've gone from zero to 6,000. It has grown so exponentially year in and year out. Even when I called my first ever game in 2021, the Next Gen Stats and the information that I was provided, the different stats we were able to provide in real time during the game on the little box on the bottom, what we can do now just pales in comparison to back then. I embrace it. It's a huge part of my preparation. I enjoy talking about it and seeing the game through that lens.

What Players Need: Film Study Transformed by Analytics

It's a huge resource nowadays for players, coaches, and obviously those of us who are responsible for presenting the game accurately. So let's put your player hat on. We're asking you to do double duty here with your player hat and your analyst hat. From your player hat, if you were going to play today, what would you ask your analytics team to give you?

It's a great question. Big picture wise, which is so easy for them to produce now, is just to get me started with understanding who this defense is. What's their play style? Are they a single high safety team? How many guys are they going to play at or near the line of scrimmage? Am I going to force press coverage or off? Is it going to be zone or man? Just give me the highlight points so that when I put on the film, instead of me having to chart every play, I know what to look for.

Back when I used to watch film, we could sort based on situations. I could pull up every play of the Dallas Cowboys' defense from the last three weeks, then filter everything by first and second down passes, zone, cover three, and forty-five plays would pop up. Nowadays I don't have to spend all that time. In one email or one phone call, the ability for them to produce everything you can possibly imagine about a particular team and strategy is incredible. Here's what they do on early downs, here's what they want to be on third down, who blitzes, is it guys on the line of scrimmage, off the line of scrimmage, defensive backs, or linebackers.

We used to have to do all of that by just the eye test, watching hours and hours of film. Nowadays these guys can evaluate plays in huge sample sizes almost in real time and produce that data in one quick, simple spreadsheet. I can tell you more about the Dallas Cowboys' defense than if I spent two months watching the film. When you get going, your body just reacts to what it's used to. That means he just watched hours and hours of film, and now he can do that a lot faster.

AWS Collaboration: Building Neural Networks and Deep Learning Solutions

So Mike, you were an analyst, and given your background as an analyst to now what you do today, which is quite different, you're now asked to give a lot more insights that are maybe a little higher level as opposed to going super deep into something. What excites you most about what AWS allows you to do? What can you do now that you wish you could have done when you were working? You used to work for the Vikings. What would you like to tell the Vikings? It's really all about the expertise that they bring to us.

AWS brings expertise that we can share, but it's not a one-way relationship. We actually get to work with the AWS Proserve team and build solutions together. That collaboration elevates our team. AWS is essentially the water that lifts the Next Gen Stats team's boat. Once we learn a technique, such as a modeling technique to build a neural net to identify what coverage a team is running, we can use that same architecture to estimate what route a receiver ran. It's all about us becoming smarter and more capable from the data science perspective. AWS has truly leveled up our capabilities.

When we started, we might have been using XGBoost and tree-based models. Now we're building deep learning neural nets and getting into the AI world. It's the evolution we've gotten to see with this whole wave of AI, going from narrow AI to more general solutions. There's so much value in the two-dimensional data we get from Zebra, the hardware service, and what we provide and analyze. We're still able to mine and derive a lot of statistics from that, such as run blocking, an area where we might be a little weak in the number of stats.

What we've found is that every project we've done with AWS—and we've done at least one machine learning project with them every year for the last nine years—we've learned something that we brought to the next project. When AWS approached us and said they would love to help us build that next series of stats, we were like, okay, we can take it on for a little bit, and maybe you can help us with that more difficult thing that is a little bit outside of our capabilities. Every year we feel more equipped with the tools, the knowledge, and the expertise to do it ourselves, and then AWS can come in and help pull off the things that we can only dream of.

Translating Data into Stories: The Broadcaster's Challenge

There's two levels to all of this. There's the data and the ability to mine it, interpret it, and produce information from it. But then there's the next level, which is the responsibility to educate people on why this data is important and how coaches and players are manipulating the data to make decisions that drive what's going on on game day and throughout the course of the week. When we get these printouts, they're much more in layman's terms. They're not from a data scientist talking in technical jargon; they're talking in football terms, which makes sense.

Our job as analysts and broadcasters is to take this very complicated data—completion percentage over expected, air yards per attempt, all these things that people are not accustomed to hearing—and speak the game of football to them in a way that makes sense. We have to continue to educate the viewer that this is not some big bad boogeyman. We are talking to you the same way they're talking in coaches meetings and in player evaluations at the combine, and how they're making personnel decisions about who to sign in free agency and who to let walk. These decisions are happening in real time in every one of the 32 buildings across the country. We need to continue to do a good job, probably a better job as an industry and as media, of talking about this in a way that's not threatening, not bizarre, and not hard to comprehend. Whether the fan likes it or not, this element of football is just going to continue to expand.

The data has to mean something. Why does it matter? We can do all the kinds of analytics in the world, but if the data doesn't matter, if it isn't driving some sort of outcome, then what's the point? We're talking about making the fans smarter, making better coaching decisions, and having players use it for their own training. It's got to have a purpose.

I think you're in a very critical role as an envoy or ambassador of the data to the fans. But there are also technologies within production that can augment the experience. We've seen this with defensive alerts, where you can start training people's eyes to recognize when a potential blitz is happening. People start wondering if it's going to happen, and then it does, and they feel smart. But you showed them that was going to happen. It's about using the data in a way that's engaging, not alienating. There's a practicality component to it.

Behind the Scenes: Preparing for Monday Night Football with Data

Sports are a nice Trojan horse for operationalizing things that people have to do in their own jobs. I'd like to make the parallel between what we're doing for sports and how someone could use it for their own jobs. Let me ask this: it's Tuesday because Monday Night Football hasn't happened yet. What does Greg ask you? Well, first we have to break down the whole team, both sides of the ball, to understand tendencies. It really starts with the coaching staff. What sport, what game this week? I have the Bengals at the Buffalo Bills. Then what will you do? We're going to look at the Buffalo Bills. We're going to start with Joe Brady's offense because that's a big story right now. Joe Brady is the offensive coordinator for the Bills. What they did on the ground, they ran the same play over and over again. I think they ran Duo exactly 100 times. We're going to look at the data and try to prove what you would see on tape, and we want to create narratives that are supported by the data. We call them insights.

We're going to gather our team of 10 Next Gen Stats analysts who are working on these packets for Greg while I'm in Vegas. It works out great. We send him an initial packet, and later in the week, around Thursday, we connect. Keegan in the audience, you should definitely ask Keegan Abdul plenty of questions. He's just as smart as I am on all this stuff. What we're able to do is feed you the insights but let you ask questions, and that's the best part. Behind our computer, we've got databases and dashboards where we can look things up on the fly. When Greg sees something, his football IQ is going to be well beyond ours. His instinct is usually right. We're going to be validating or invalidating his thoughts or what he sees subjectively. That's what we do on a day-to-day basis.

Friday night, we got into a whole big thing on the two-point conversion decision. Greg had a lot of conversations about this decision on Twitter, and a lot of that was fueled by our team. We were able to do that research and were emailing in real time. I was copying and pasting it, so we're able to provide that support ad hoc, whether it's for the broadcast or elsewhere. We only do that for media. It's really about making the fans at home smarter, bringing leveling up the game, and making them feel as close to the coaches in the booth as Greg is.

Let me recap this because we have international audiences and a lot of different people. Duo is basically running to the inside. Let's make it very simple. Greg watches the game, maybe he's paying attention to Seattle because that's the game he was on this week. He's watching a little bit of that and he's like, wow, I see them run the same play over and over again. Then he can say to the team, I think they just ran a lot of Duo. Is it because the Steelers, who were playing Buffalo, are bad at defending Duo and they just didn't adapt, or is it because they were doing something special with their offense? They can go into a lot of depth. Sometimes Mike comes up with things that Greg's like, oh, I didn't know it was actually this, or maybe I should look at this. It helps direct the traffic of what's going on. The data insights are a collaboration between the expert eye of a guy who did it for as many years and at such a high level as Greg and also what AWS and Next Gen Stats can help illuminate that maybe you didn't miss because this person isn't even watching that game.

The Two-Way Street: Validating Observations with Analytics

Ultimately, all of those different things work hand in glove, so you're getting a better storytelling situation. The people who benefit are all of us because when we're watching whatever happens in this Bills and Bengals game, I'm pretty sure the Bengals have a really bad record. They allow the second most yards to running backs. I know that. So you know it's going to be interesting. You're exactly right, and it really works both ways because I get the initial packet from them early on. We have our call later in the week. We're actually on a Zoom and just having open dialogue. They're in real time. Yes, you're right, or there are often times where I say, hey, after watching them all week and reading all this, I'm seeing this. Sometimes their data will validate what I observe, and sometimes they'll go, you might think you're seeing that, but it's not.

Now I know I don't have to get on air and be as emphatic about making a clear point because maybe my observation isn't backed up. And then vice versa—I'm watching the film after reviewing their initial packet, which breaks down and gives me some general points. When I'm now watching the film of last week's game, I have all these preexisting storylines and trends already in my brain because I've read their packet, and now I'm watching the game with a clearer context.

I would have to spend almost unlimited hours to find out what coverages they run on first down, what coverages they run on second down, what coverages they run when they throw the ball on first and ten, or if it's an incomplete pass, whether the next play is a pass or a run. It would be completely impossible for me to watch every play, chart it all, memorize it, and be able to bring unique insights to the viewer that go beyond just saying "nice throw, nice catch, he's a great player, he makes fifty million, and he went to school at Georgia."

We're trying to go deeper with the audience than just narrate what's happening. We tell people why it's happening, what is going to happen, and what are all the conversations leading up to what we see happen on game day. There's a reason for all of it. The last level to that, which is probably what I rely on them for during the game the most, is it could be the simplest thing—like, on the last five first downs, they've run the ball. That stat to the audience could be a lot more interesting if it's presented as: on the last five first downs, the Buffalo Bills have run the ball. They passed the ball at a ninety-five percent rate, which is the second highest in the league. That is very contrary. This is abnormal. Why are they doing it?

Well, it's because the Cincinnati Bengals have certain characteristics, and then you can start getting into the cause and effect as opposed to just saying the run game's been really good or they're very run heavy on first down. Are they always run heavy? Is it because the Bengals stink? Is it because of them? Is it because they're winning? Is it because of the weather? Tell us there's more to it than just telling people what's happening. We need to tell them why it's happening, and their data allows us to speak very confidently that what we're sharing with the audience is true.

Beyond the Game: Player Health, Safety, and the Dynamic Kickoff Rule

That actually segues perfectly into some of the evolution of Next Gen Stats and what we're able to put on top of our initial Next Gen Stats because I think what's being understood is why should you care. Why do I care if the Bills are running three times or five times on first down? Well, it's because they usually pass on first down. All of these things help us develop better Next Gen Stats, deeper and better things built on top. For context, Next Gen Stats is about five hundred million points of data over the course of the season. That same data is an input to a whole other swath of data, which is also five hundred million points of data, but on a weekly basis informing player health and safety for the league.

All of this speed, time, and distance feeds in. But now with optical tracking and cameras, we can also capture skeletal pose and understand where players are maybe incurring injury or at risk of injury. The league made a player health and safety portal available to all thirty-two clubs two years ago. It was the first year that seven hundred fewer missed games by players occurred when that rolled out. They're able to understand which plays are more injurious than others. The swivel hip drop tackle caused a twenty times injury rate over a normal pass or run play.

The dynamic kickoff rule changed, and people were all up in arms about it. But let me tell you, we were trying to solve for two things there: a two times the injury rate and four times the concussion rate, and nobody was actually returning the ball. So you had an uninteresting play that was far more injurious than a normal pass or run play. Taking ten thousand seasons' worth of data, the league simulated the rule change to optimize for those two things: higher returns and lower injury rate. So you saw that unveiled last season, and it's here this year as well because it did exactly that. That rule change brought the injury rate down to a normal pass and run play, and as of week nine, there was a seventy-nine percent higher kickoff return rate. That's just one example of using this data and other data to look at the data in a different way and find other insights. Equipment has changed, rules have changed.

And you're using the data in other really fun, interesting ways. You'll probably talk about the Fantasy AI Assistant using generative AI and agents to help fans draft better and make their lineups every season. So it's always something interesting, but at the base of this, it's making sense out of really large volumes of complex data, and I think that's something that transcends every industry.

AI as Augmentation: The Fantasy AI Assistant and Research Editor

That's exactly right. Just to piggyback on that with the Fantasy AI Assistant, I'll actually be on stage tomorrow at one o'clock giving a talk on the Fantasy AI Assistant that AWS helped build for us a few months ago, and it's just one of the many products that Next Gen Stats helps power. You can imagine that we can create as many stats as you want and feed them to Greg in the broadcast, and you can see it on TV while you're watching the games. But what about those products across NFL.com, across NFL Plus, what you see across different platforms, whether it's video games or AI?

What we're really getting closer to is how can we take advantage of this new wave of technology. With the Fantasy AI Assistant, we decided that we've got a lot of great data. We're sitting on tons of great data. We've got Next Gen Stats data to add that context. What if we were able to create this Fantasy AI Assistant? We had the data for it. We didn't have the expertise in AI, and that's where AWS came in to level up our ability.

We were able to not just develop a Fantasy AI Assistant. We've developed a research editor. We have a team of ten analysts that are at our office right now preparing for week fourteen. What they're doing is writing three to four sentences on every player in every team on every stat we could possibly come up with. But what if their job wasn't about writing it in the perfect format? What if their job was to just find the best research? We created an AI assistant that helped edit their work. Before Greg even gets his packet, we made sure that the style, the format, the tone, the stats, everything went through the system.

But what we didn't want to do is say, "Hey AI, go do the research, go spit it out, and we'll send whatever you have." We didn't trust it. Now as you see AI get better over the years, we are sort of catching that wave of, okay, let's first do it mainly from the human researcher side. Let our analysts do the research and help them. Right now we're going to get more into augmenting their research. The ability to find things faster, to augment what they write in a way that doesn't just replace them but allows them to do their work two times as fast.

It's not about cost saving to replace analysts because you need that human context. You need to know that Lou Anarumo was the defensive coordinator of the Bengals, and when he left, the whole Bengals defense went to hell. Maybe they shouldn't have fired him. You can't just trust an AI to pull that context. But what if you have the ability to serve the human analysts to use these tools to two times their time or four times their time? They can not just put out fifty insights but maybe one hundred insights a week. That's how we see the future of AI happening, not as a replacement to human analysts, but as a supplement to their ability to do their job faster and more optimally.

All of the things that was really well said. Yeah, absolutely. So I don't know what he's talking about, but it sounds great and wonderful, right? He's killing it. I take back all the nice things I've said about Next Gen Stats. I should just be thanking the AI bots. Well, no, they're just because I didn't realize all this work that I get each week. These guys are in their office grinding. They're in Vegas and apparently just running machines, so I take back everything I've said nice about you guys.

The Big Data Bowl: Crowdsourcing Innovation and Creating Career Pathways

Well, I've got something that you can use if you want to that I was thinking about with player health and safety. I heard it from a coach. You can borrow it as you want. The reason that the guys who are the replacement players aren't the starters is because if they were better than the starter, they'd be starting. You can work that into your analysis this week if you want to. That actually does bring us into something that has been a genesis for thought for our whole team, which is the Big Data Bowl.

It is a very special event. If you don't already know what it is, it's an annual competition. There's actually money you can win. One hundred thousand in prizes combined. You get to come to the Combine, which if you are a football fan, truly I actually think that's worth more than the money because you get to see all these people walking around in their team gear and they look great and everyone's happy because they think their team can win the Super Bowl. Ultimately, what the Big Data Bowl is, we open up a certain amount of the data, the Next Gen Stats, to teams. There are undergraduate teams and there are graduate teams, and we say solve a problem around a topic. We create a theme every year.

We get to hear from some of the smartest people who come up with the coolest things that have actually turned into real parts of our Next Gen Stats analysis. You are looking at two of the three judges. You're not a judge, but a prejudge. I have the first few. That's right, so you help refine it before it gets to the judges. Let's talk a little bit about the Big Data Bowl. It is the eighth year, and we're very happy about that. The COVID one was challenging, but now we're glad to be back in person.

What has come out of the Big Data Bowl, and what is the best part from your perspective? Well, it's threefold. It's a win-win-win. The NFL wins in terms of the ideas that come out of it, and we get to build new stats. But really, the people who participate in the Big Data Bowl are getting direct access to these teams, and their ideas are fueling our ideas and what we're working on. When we release the data to the fans, it's about how we can leverage the expertise of the world versus our own team, and it's really all about inspiring the next generation of stats.

We've created route classification, coverage classification, tackle probability, and pressure probability. All of these stats came from fans at home working with our data. They didn't even have to be football fans to come up with these solutions. They just saw a problem and saw that it needed a solution, and they applied their expertise to it. The 2020 winners of the Big Data Bowl had a model where the difference between the first place team and the second place team was the same as the difference between the second place team and the twenty-seventh place team. There were two Austrian data scientists, and they had no idea what football was.

They saw it like a video game, similar to Pong. This person has to get to this area with things in the way. What did they do? They created a neural net, specifically an LSTM. We still use that architecture today. That solution is now the architecture for several of our models. Because of that, we get to leverage the expertise of the fans at home. If you're interested, if you're a data scientist, if you're a football fan, find a data scientist and help create that next metric that we certainly miss and certainly need.

I've judged a couple of years now, and I'm excited to do it again this year. I love it because there's so much enthusiasm around it. But the interesting twist this year, which I think is super important, is that it's not just the data science behind it. There are some really tremendous and impressive analytical work being done, but it goes back to the point of why does this matter and how do we tell the story about it? That whole layer matters. Last year, the prompt was pre-snap tendencies. I don't know the prompt yet this year, but it's not only solving the problem or answering the question, it's then how are we going to communicate this to people so that it matters and that they will care?

Making Analytics Accessible: Storytelling for Fans and Growing the Industry

That matters not just to the hardcore fans, but if you look at the league, they had seven games internationally this season, and they're not slowing that down. The league has this unique challenge of having to explain a very American sport to the rest of the world. We're talking about these analytics to the hardcore fans, but you can turn this all around and introduce this game to new fans in a way that they're going to understand it if you actually tell the story correctly.

Absolutely, and I think the biggest thing is processing what's being fed but also making sure you're doing it in a very approachable way for the fan. That's the biggest challenge that we have calling the games. What we get pumped to us as far as the raw data and the trends and completion percentage over expected, we can't get on air and start saying those things because people have no idea what we're talking about. But we can present the stories, we can present the context, the trends. We can use the real, nitty-gritty data points as just the foundation and the backbone when we tell the story about what the Buffalo Bills' offense is and what they do and their style. We can talk about why the Philadelphia Eagles on 3rd and 6 when they break the huddle run the ball at a certain percentage, which is the most in the league and it's over expected in the league average.

We're never going to get that across, but just a simple story between the two broadcasters on ready break as they come up to 3rd and 6 and saying, hey, throughout the whole league, you look around the league on any given week, Joe, they're dropping back to pass. They've got multiple receivers on the field. Not the Philadelphia Eagles. They are going to hand the ball off, mostly from the gun. Why are they doing it? Because they're setting up 4th down and it gets into the tush push. It gets into like there is a nitty gritty football strategy component, but we're not beating people over the head with run plays, pass plays over expected, and you're giving percentages and averages and rankings and people are looking at you with glassy eyes.

So you can tell the story that the data supports in a very easy, fun, light way that people go, oh my God, I've been watching Philadelphia games all season long. I always just thought everybody around the league ran the ball like this on 3rd and 6. I didn't realize this was so abnormal. I can't wait to watch more going forward because I'm going to be curious what they do. That's a fun moment. That is data supported, but it's not just regurgitating the data on the broadcast.

But I'll tell you what that does though is that if you can make a fan feel smart about what they're watching, they will watch more of it, of course, right? And so that's the whole point and that's our job, having people like, oh, I am interested. I'm going to pay attention more. My antennas are up on this, right? That is a more engaged fan, which the league wants.

The trick of it though, not to belabor the point, the trick of it is fans have been taught the same thing for 30 years. Week in, week out and you just hear every coach, every broadcaster, every player just regurgitate the same couple coaching points in the press conferences, in the pregame interviews, the postgame interviews, in the locker room, and on air. Broadcasters still are guilty of doing this where they, to your point, you know what fans are used to hearing, so you almost find you have to be careful that you don't just speak to what the fans have been told like, hey, coming into this game, you know, the team that works the hardest and really comes out, starts fast, and they really gotta make sure they stop the run and you give all these generic clichés and what are those fans doing at home? They're going, I knew that. I knew that. I've heard that before. I told you last night at dinner I told you they needed to stop the run.

But it doesn't mean that it's true, so there's this weird go-between. We gotta make sure we educate and speak and connect directly to the fan of what's actually happening, but then you also realize if you just tell the fan what they already know, they're going to go, this guy Mike, he's pretty smart. He tells me everything that I already know. I like this guy. So there's this weird balance between the fan and the viewer and the broadcaster.

Fans don't like being told things they're not comfortable with, or they've heard the contrary for so long. You don't like my team. You don't like this player, you're a hater, you like the other team. No, I'm just telling you what's happening. They don't, so there's this interesting back and forth. There's nuance to that, right, because yes, they want to feel smart so you can tell them what they already know, but then everybody likes to have the one thing that educates them, and in a way that like, oh, now I know something else, and you're not speaking down 100 percent.

And there's also the advantage that we have in football over the other major sports, and that is the style of the game. There's a lot of advantages to having stoppages. Out of a 3 hour game window, there's only about 15 to 16 minutes of actual play. That gives us about 2.5 hours to be able to teach the fan at home. And not only is there a game, you look at basketball, and it's a little bit a refined game now. It's dunks, it's 3 pointers, it's fast pace and tempo.

Baseball, it's homers, strikeouts and optimizing fielding in your pitching arsenal. Well, in football we've got games within games happening all over the field. You can analyze the left tackle versus the defensive end. You can analyze the receiver versus the cornerback. You can analyze the whole coverage unit versus the route combinations. There's so many games within a game that you get to zoom out and zoom in on that we are so lucky to be able to have so many storytelling devices to be able to fill that 2.5 hours with really compelling insights.

So before we move totally off the big data bowl, I do want to point out my favorite thing about the big data bowl, which is company sponsored health insurance, also known as people getting jobs, because I think that's a very important thing. I love that people were at over 60.

That's right, there are now over 60 current NFL staffers with teams today that get health insurance who were formerly participants of the Big Data Bowl. So in 8 years, you can do the math. It comes out to roughly 8 per year. This is the pipeline for analytics staffers. We have teams reach out to us all the time. We've got VPs now in analytics departments across the league, and they thank us because not only did we provide them a new data point through the event, but we provided them with more talent to fuel their very big and growing departments.

That's something that has changed. Talking about the evolution of all of this, you used to have a couple of teams maybe have somebody seasonal or interns doing some work. But now these analytics staffs are among the biggest staffs on teams. They're investing in this, and it's not just from a game strategy and coaching perspective. Players are asking for this data too. Agencies now have access to all of the surrounding industry of it. I think it's important to point that out because I like when people have a chance to grow their career.

Q&A: Real-Time Graphics, Business Applications, and the Future of Analytics

So we're going to open it up to you guys for questions. Listen to what we're saying, but also get your questions ready. Mike, we'll start with you. What are you most excited about with what's currently available? Give me your current favorite and then your dream scenario stat, whatever it is.

I am an AI optimist in the sense that I've already felt the productivity boost personally. It's all about how we can bring the tools, the products, and the data to the fans and internally, spread it across the departments at the NFL where we can create not just narrow AI to know what coverage a team ran, but more general AI to really be able to ask any NFL question. We would love to create an AI bot or an AI assistant that had access to the NFL record book so you could ask any questions about the history of the NFL.

We would love to set up an AI assistant that could help with orders. Just give me one that you want. I would love to get to the point where we can output AI-generated graphics that we can feed directly to the Fox broadcast so we can go from our research to graphics. Throw that graphic back up there. We should do that. You can just use Nova for that. Exactly. We're going to, I think we just came up with new products. So this is one of the current graphics right now, graphically represented.

What you're looking at here is the amount of under pressure. Greg, you take us through this. So under pressure again, this is the first time I'm seeing this, but in the first half, obviously self-explanatory. He was 4 for 6 and 83 yards only while facing pressure. Pressure is when the defender is within a certain amount of distance at the time of the pass. When the defender is within a certain distance of the quarterback at the release point of the pass. Compared to a blitz, right, so people always say they pressure a lot. A blitz is when you bring 5 or more rushers, more than 4 rushers. That is a blitz. A pressure is just how close you get to the quarterback. When you get into that halo that's overlaid around Jalen Hurts, and then the rest of it is obviously clear.

In the second half, according to this, he was pressured 4 times. The 4 passes when he was pressured, he did not complete any of them. These stats are more like traditional stats just under the guise of pressure versus clean pocket, which would be the contrary. All the rest of the throws would be that. How we got to this, we've done this throughout this whole talk where we are playing a game of telephone with Greg. We want to tell him the need-to-know football perspective of what we're seeing. But this under pressure value, we have a time series for every pass rusher, the probability of generating a pressure on a play, and that is how we generate and identify what pressure. But we don't want to tell Greg that. We just want to tell Greg, hey, it's when the quarterback's under duress so that he can tell the story of, hey, Jalen Hurts really struggling under duress in the second half. Obviously they lost both games to the Cowboys this year, so it definitely played a factor.

It would be amazing if, as we said that on the broadcast, it was processed. You guys are manually listening to what we say. The issue with the lag time is that certain crews have different access to these guys than others. There are some steps in between. If you pay us more at the NFL, we'll give you access too, just so you know. That's a conversation for another time. Certain crews have more resources than others. We'll just leave it at that.

These graphics are only as good as their ability to reflect real time. At the time the announcer says something, the shortest amount of time that passes before that comment is supported through a graphic like these that reinforces what the commentator just said. By the time the commentator is heard, the comment is processed, the graphic is created, sent back to the truck, and then broadcast on screen, if you've been under pressure two more times since that happened because it took a minute and a half, the graphic is completely irrelevant.

The real-time communication between the booth and the people generating this information is critical. The tighter we can shrink that lag time, the more willing and able we are to have these types of graphics support the words and pictures the broadcasters are sharing. For everyone listening, I think this is a challenge. AWS can help with that. Here's the thing: we get closed captioning in real time. Why don't we just read your closed captioning data and have one of your smart AWS engineers create something that could read that closed captioning? I'm going to tackle that. There you go. I think that's something I can handle.

Mike is saying it's just the analytic itself. We know it's happening. The graphic is generating in real time, so it's almost going the other way. The graphics are wonderful, but what we have to do is work with your crew during the offseason to create the API that would integrate into that specific graphic. If we were to create that, then we could get it as soon as Greg says it, and the under pressure graphic could go out. We can inch our way to getting more and more of this data in real time to the fans.

Look at the one on the left about Caleb Williams. Just as a viewer, how much more interesting is a conversation around Caleb Williams when you're talking about how last year the Bears were historically bad taking sacks? Everyone said the offensive line stinks, he stinks. When you're talking about Caleb Williams, how much cooler is it to talk about his improvement through the context of this? The number one value of a quarterback is when you're under pressure, don't turn the ball over, don't take sacks, and generate big plays on the move. If you look at the top quarterbacks in the league, you can have that kind of high-level conversation about Caleb Williams.

This graphic gets put on screen through the data they have to further reinforce the story. You can't say all of that in a sentence like "Hey, this year he's been sacked 119 times. He's only been taken down on a sack 14 times, so it's 11.8%. Last year he was 17%." But there's a cool story about the improvement of Caleb Williams and the improvement under Ben Johnson and the improvement of the Bears. You can have that conversation and then let the graphics speak about the real nitty-gritty stuff, which is essentially what my comments, while more general, are based on me spending all week reading that information. I'm just generalizing it into a more conversational tone rather than just beating people over the head with numbers.

You could also get into a conversation about strategy because sometimes games get boring at the end. If it gets boring, I called a game yesterday that was 26 to 0. Sometimes that happens. But ultimately, they get a little boring at the end, right? So you put that graphic up and pretend it was the Bears in this case, up 26 to 0. You say, all right, well, what's Ben Johnson doing? You could have a great conversation about different run plays, different ways to get out of trouble, whatever it is, and you could start a conversation from that as well. It's not just spending the whole time talking about the ball. Nice throw, nice catch. He ran fast, he made the guy miss. We see that. If you don't do this, it just gets very redundant and boring. In my opinion, that's boring.

Does anyone have a question ready? The analytics. In the spirit of getting people healthcare and making analytics, making the sport safer, do we see a path to applying what we are doing from the NFL at a lower level of football?

It's already happening. Look at the equipment. It used to be that everybody wore the same kind of helmet, right? But the offensive line is going to absorb hits differently than a quarterback versus a wide receiver. People are taking hits in this game much differently. So now you have position-specific helmets that didn't exist years ago, and that has trickled down to college and even the high school level. So the investments that are being made, and I would argue that if you guys are banning hip drop tackles, everybody else is banning hip drop tackles. The investments that are being made, especially in this player health and safety side of the NFL, have impact not only within the game of football but all sports.

I also think the pressure of the league investing that much in player health and safety should put every other property on notice. You should be doing this too, and I particularly would love to see the same level of investment and research and rigor applied to women's sports and health and safety in that space as well, because that is a very barren space right now. If you go check out that exhibit, there's a series of helmets over there. That's probably the biggest difference. Certainly what you could do with our Next Gen Stats tracking data is measure collisions. We're not doing that at Next Gen Stats. We're not trying to tell stories about collisions, but you can bet the player health and safety department is measuring and trying to identify what is leading to these injuries and how we can make the game safer.

The helmet is something I see with my two middle school boys. I coached the middle school team at their school. The helmets that every single one of our boys wears—not just my two boys because we can get one, but every single kid on the team, there are 50 boys on the team in 7th and 8th grade—every single one of those kids wears a helmet that is the exact same helmet when you turn on TVs on Sunday that the NFL guys wear, just in different sizes obviously.

What can business leaders learn from all this? I can take that one for you. What I'm looking at across different businesses, the thing that people always ask me about Next Gen Stats is how do I speak clearly? What is the communication? Because it's a two-way street. Greg needs to be able to ask the people who know how to do the technical part in a way that they can actually understand what he's asking for, and they need to communicate to Greg what the insights are and what the deliverables are for your company. It could be which products should we focus more of our money and resources on versus others.

It's the communication. I need to say if I'm Mike, I need to say these are the parameters of what we can do. This is what we're able to give you. Greg needs to say, I'd like to know this, and then you say, here's this. It's that give and take, the communication between both of those people that makes the whole thing work. Because if he doesn't know how to ask the question and he doesn't know how to tell him what the answers are, it doesn't really matter if you have the best data in the entire world. No one's going to use it.

I'll just add to that. I think it's the investment in it, right? What the league has invested an awful lot in trying to make sense of their data in a lot of different ways. We're talking about largely Next Gen Stats. We're talking about player health and safety. They've also invested a tremendous amount in understanding their fan data a lot better. They had 90 billion rows of fan data that they had to try to structure and get insights out of. So I think it's just that mindset.

What do business leaders need to do? Every business has data. Are you investing in structuring that data, trying to glean insights from that data, being predictive about your modeling with that data to help you prepare for the future of your business? So I think it's all of those things and admit what you don't know. That's the other one that I found. Because I think a lot of people are like it's perfect and then you're like, no, your model has this flaw and then everyone moves on.

So to Greg's point about talking in storytelling rather than in data or data storytelling in some ways, but the idea that you have to communicate the need to know things for your business, right? It's about putting it in a way that you could distill. I've never mentioned a neural network to Greg before, right, but I've mentioned it here because I understand that there's probably some technicals in the audience that want to hear about how the sausage is made. When we talk to Greg, we're talking football. So I think it's all about how do you leverage the information and the data you have to become experts in your data, but then to be decisive and to know what's valuable, to know the context.

Once you're an expert in your own data, then you become more confident that you can make those data-driven decisions that do come with that positive expected value out of the outcome. Do we have a right here? He's got the mic. Someone with the mic, go ahead, whoever has the mic, and then we'll get to you next.

The Human Element: Why Data Will Never Replace Coaches and Monday Night Predictions

I have a question for maybe Cynthia or Mike. Do you guys think that the analytics will mature to the point where it can essentially replace play callers? Hopefully not. Absolutely not. No, absolutely not. There's always that we get this question a lot too, which is the data is taking over and the data is making decisions. No, it's not, because a coach is going to know if he tweaked his ankle on that last play, if he's got something going on in his head, if the locker room wasn't right, there's something going on. It is a tool in the toolkit. It is not the answer.

I remember ten years ago at the MIT Sloan Sports Conference, Billy Beane was up there with his two deputies. I think it was Paul DePodesta who was with him in Oakland, and then the other GM who was with the Dodgers at the time. It was the most fascinating talk I had heard because obviously I'm born out of the Moneyball era. What Billy Beane said blew my mind. He said that early on, I think it actually might have been Paul DePodesta, who's played by Jonah Hill in the movie if you've seen it. He said when we first were printing out spreadsheets and giving it to Billy, we were basically giving it to him, but we're pulling it back because we still wanted Billy to use his intuition and his expertise that he's developed.

When Billy said that he doesn't make a decision any higher than thirty percent analytics, seventy percent gut. That seventy percent gut is trained over twenty to thirty years, and the data should either support, validate, or check whatever it is. The data is just a tool, and there's so much context that it doesn't measure. How does the left tackle play? Is there an injury that you need to know about? Is there some psychological bias or advantage that you can feel that the data won't feel? When it comes down to it, the tools will feed into the future play callers to make them more optimal decision makers than just an automated decision alone.

When I worked for a very famous coach, he told me that he would like to know his own biases so he doesn't make mistakes. He told me that he believes in his system an average player could be above average if the right attributes are identified and utilized. There were very clear parameters of what this person wanted and what this person didn't want, what ills they could live with and what they couldn't in play calling and in all things. What he was trying to do very smartly was not allow his own internal psychology to bias what he was calling on the field.

It's more of a check and balance and more of a tool in the toolbelt because we don't really have a hard time identifying very elite, good, and very not great. The middle is very messy. So how can I select for attributes that make my middle a little bit better than worse? Thank you for your presentation today. All of the information is really interesting. There's so much data that you deal with, but I'm wondering, is there a list of data you wish you had? Like what don't you have yet that you want?

I could actually say that we should not give him an answer here because he will be here for twenty minutes. The data that we're actually now collecting for the first year, we're going from two-D to three-D. This is the first year we've collected it through Sony Hawkeye Systems. They're in soccer, they're in tennis, they're in all the different sports already. What this is going to unlock is full pose estimation, twenty-seven point full pose estimation at sixty hertz per second.

What that's going to do is we've got a couple more stats to do with two-D and then it's going to be a whole new inflection point about what we can do. Angles, release point, stance, just knowing if a player has their hand in the ground or if they're in a two-point stance or they're route running. The amount of data that's going to come out of this three-dimensional data source is going to not just affect the stats that we can create, but it's going to affect the way that you watch games at home.

Next Monday night on ESPN on Disney Plus, there is a Monsters Inc. game that will be featured for the Eagles and the Rams. This is using our data in near real time to recreate the players on the field using Monsters Inc. avatars. If you have kids at home, please try to turn that game on because it's for them. The whole goal is to bring data not just in the form of stats but through experiences, and we're excited for the future of the fan at home and what they're going to be able to see.

Not just during TV broadcasts and not just all broadcasts, but online through applications and products and tools like NFL Pro. This is the inflection point of the next era of Next Gen Stats. How many monsters and characters can you name? That's really the only question that people ask. I have a one-and-a-half-year-old, so I'm just outside of the realm of the monster sing. You can make AI software and predictive algorithms, but if you don't know the characters in Monsters Inc., I'm just going with my favorites: Mike, Sully, Terry and Terry, Bob. How about Boo? I know them all. I've only seen that movie 700 times.

We have two more questions here. My question is for Mike. I'm not sure when you worked for the Vikings, but I was one, so I don't know if the data availability is comparable. Could you compare the kinds of questions and models you're working on with the team versus what you've done with Next Gen Stats or how it's changed? That's a great question. Ten years ago, we were using Microsoft Excel. R was becoming a thing as a programming language, and Python was just in the early days of data science. We were taking all of the data we had and trying to work with it. We actually hired a data scientist to help us with the 2015 draft. That was the draft where we ended up with Stefon Diggs in the fifth round and Daniel Hunter in the third round. It was just a tool, and they aren't on the team anymore. Well, they're still in the league. No, they're in the league, but they shouldn't have left.

The idea was how can we help the scouts make better, more informed decisions. So what we did was we said, how do we make an impact on the draft with the results of our models? We just said, if you're in the top twentieth percentile, you probably have a better chance to make it than the bottom twenty-fifth percentile. So let's just put a blue sticker on the players who are in the top twentieth percentile and an orange sticker on the bottom, and we're not going to do anything else. Rick Spielman was the GM at the time, and he just went ahead and used that as he may. He ended up drafting nine blue stickers out of our ten picks and zero orange dots, and it ended up being a really strong draft class. It was just about saying, your scouts set the board, now let analytics validate your decision, and that's how you make decisions.

Now they have more data than you can count at the college level to make much more robust models. You're no longer living in Excel, and you're now in Python and getting much deeper into the architecture to predict performance. I think the amount of data, the capabilities, and the size of the staff has really made it so that the NFL draft isn't seeing suboptimal decisions as you were ten years ago across the board. What's the story around providing some of this data to individual teams? I can answer that. They all get access to the same thing, but it's gated at certain times, so they don't get it during the game. The idea from the competition committee and the football operations department was that during the game itself, we want to limit the amount of technology. We wanted to make it so that you can prepare as much as you want, but between the whistles, there's no technology involved. So our internet is off, no internet at all.

If you have something, you can use Excel. That's actually a new initiative that we started last year. We call it Spreadsheet in the Booth. There's one computer with a spreadsheet that you can program ahead of time. No internet, no Next Gen Stats, no RSS feed, no CSV file, zero new stuff. The idea of competitive balance is key and crucial. Now you've seen teams certainly ingesting the data as soon as the game is over, and they're now creating those reports for the coaching staff. Now you know that the players hit their max speeds or how they're playing relative to their baselines and stats. It's really about player health. They get all of the same information. It's gated at a certain amount of time, and how they use it is where you can get an edge.

Teams who have done a better job either investing in the right people or have the smarter people do it better. Teams that have not, I'm not sure if they use it. I'm not sure, but they get it, right? Some of that position data would be awesome for position coaches, right? They get it if they want it. It's all there for them if they can get it, just not during the game. Alright, so my last question is because I don't get a chance to ask Greg Olsen to give me personalized Monday Night Football predictions about who's going to win.

We get the best analysts in the business, and I'm not saying that because you're there. I say it all the time when you're not around, so I will keep saying it. We can negotiate on that. Let me phrase it this way because everyone will ask who's going to win and why. I don't want to know that right now. The Giants are underdogs, we know that. We know there's some history there, maybe like a good helmet catch. How did the Giants upset the Patriots in this matchup?

Well, obviously I think the return of Jackson Dart. Jameis Winston is fun, and he is a drill a minute. He should be permanent in the third quarter just with a camera and a microphone. I saw someone say he's like the universal backup. The second you lose your quarterback, you have to play Jameis, and he just plays for whatever team and does whatever he wants. I audibly laughed when someone said that because it was very true.

Jackson Dart is back. This will be the first time that Jackson Dart has played quarterback for the Giants with Mike Kafka and not Brian Daboll. So what does that look like? I don't think anybody knows. I don't think the AI knows, the football, we certainly don't know. We called the last Giants game, which ended up being the last game of Brian Daboll's tenure. It was in Chicago and they had the big lead in the second half and they decided not to go for the touchdown on fourth and goal at that half yard line. They kicked the field goal, and Caleb Williams took them back to back for touchdown drives and they lost. He got fired a couple of days later.

I don't think anyone really knows what Mike Kafka's vision looks like. What does a Mike Kafka-led team, led offense, led locker room look like with Jackson Dart under center? They've also since fired their defensive coordinator, so that's a whole new thing. Things are going great. I think there's a lot of unknowns for the Giants. Would I be shocked if all of a sudden they went out there and beat New England tonight? No. I think New England is good. I think Drake May is fantastic. I think he's that next group of superstar high-level quarterbacks for the next ten-plus years. I think he's that type of physical player.

He's got that sort of mentality and approach. He's been like that. He grew up in Charlotte. I've known him since he was a young kid. He is special. He's unique. Do I think that Patriots roster is a Super Bowl roster? I don't. I think Mike Vrabel has done an unbelievable job turning that around. They and the Minnesota Vikings spent more money than anybody in NFL free agency, certainly under the idea that you have a young rookie quarterback, Drake May, and respectively, JJ McCarthy for Minnesota.

You spend all that money in free agency on veterans because you're saying I have a five to four-year window where I'm not paying my quarterback sixty-plus million dollars and I can allocate my resources. It appears as of now like the Patriots struck gold with these contracts, this roster turnover, obviously drafting Drake May, hiring Vrabel. It's been one good decision after another, and they're the number one seed if the season ended today in the NFC. On the best team, again, everything we know, New England is the best team. I think as we see every single week, sometimes one little flip, one little change, the game goes your way. Would I be shocked if Jackson Dart comes out tonight, plays free, plays fast, operates this Kafka system and they beat New England? I wouldn't be shocked.

Mike Kafka is a former quarterback. Do you think that changes anything just from like, you're a former player, put on that hat for a second? Brian Daboll had the experience with Josh Allen, of course that's kind of how he got his job. Mike Kafka comes from the Patrick Mahomes Chiefs, and he was a quarterback in college. He played and had a cup of tea in the NFL. How do you think that sort of changed things? I think the biggest thing that at least I'm looking for is hope for some job. Maybe there's no, again, I just call it as I see it. I've got a good, I particularly care who wins, but I'm just calling it balls and strikes.

I think what's going to be really interesting is Brian Daboll's history with Josh Allen early in his career and even yesterday. Josh Allen has always been a run quarterback. There's a run element. Jackson Dart's some of his best plays this season have been the design quarterback runs, and then all of a sudden in that Bears game, he ran for two touchdowns on designed quarterback runs. On a third and long run, he gets blasted, fumbles, smashes his head into the Soldier Field turf, gets a concussion, hasn't played since.

So now he returns. You're an interim head coach theoretically coaching for the job. I don't know how realistic whether he gets the job or not, but you're saying I want to win or else someone else is going to coach Jackson Dart. You just lost him to injury a couple of weeks ago running him. Can you be a good offense if you don't run him? Can you win?

But is the front office saying we can't get him hurt, the season's over, and then I'm Mike Kafka going, well, my season's not over. If we stink, I'm looking for a new job next year, and yes, I don't want Jackson Dart to get hurt, certainly, but protecting the future quarterback of the team that I'm not the coach of is not in my own best interest. So I think there's a lot of very interesting aspects at play for interim coaches and young quarterbacks last season.

If you were going to give the one stat for tonight that you think is the most predictive, just one, what's the most predictive one from the Next Gen Stats packet? Brian Burns and his ability to create quick pressures. The edge rusher for the Panthers—just to provide context, they've got Abdul Carter, the rookie that they just drafted last year, and then they got Kayvon Thibodeaux. So they've got three really good pass rushers. Lawrence sometimes rushes the passer too. His pressure numbers are down a little bit this year, but he's always one of the best nose tackles.

What Brian Burns brings to them is the ability to create quick pressure. If you get to Drake May, he has been prone to throw interceptions in his career going back to UNC. The X factor is Will Campbell, the third overall pick last year, the left tackle for the Patriots, who went on IR last week. He had played every game and had been lights out at left tackle. If the Panthers can get their pass rushers going—Brian Burns, Carter, Thibodeaux, and Lawrence—and they could take advantage of a little bit of a weak link on that left side of that offensive line, it could be a difference maker to create those negative plays that would keep New York in the game for the long run.

I've promised the Giants defense that I would make fun of them about how bad they are at defending outside runs until they stop doing it, and they won't stop doing it. If the Patriots are going to win, you're going to see some nice outside rushes because the Giants are historically bad. The highest explosive play rate on outside runs allowed is the Giants by like a million. The most yards on any metric, any Next Gen Stats that you want on outside runs—who's the worst? The Giants. When you rush those outside edge rushers, you give up containment on the outside. So Traveon Henderson and Rhamondre Stevenson could certainly take advantage of that.

We're not talking about the T word today because we have a Michigan grad here and he went to that terrible school down south of Michigan, so we're not going to talk about him. Don't worry, it's four to one over the past five years. It's okay. So the Patriots are my pick. I'm also a Patriots fan, so go Pats. I wish the Lions were better this year though. They were ten and zero.

You can't account for strip sacks and interceptions and pick sixes. That's what makes predicting these games so hard. I always start off my evaluation saying an even game, an even possession offense versus offense, not taking into account punt return touchdowns. Who's better, right? And now all of a sudden when the better team gets something like that, it could be a long night. It's virtually impossible for the underdog to overcome that unless they get a bunch of them on their side. You look at the Panthers beating the Rams yesterday. Stafford hadn't thrown an interception in what seemed like a year. He throws one on the five yard line and then the next possession he throws one for a pick six. You can't bake that into your predictions. You don't think that's going to happen, but it's the game.

Well, thank you all so much. We held you over because this was such a good conversation. Thank you guys so much. Same watch Monday Night Football. There's tons of great conversations here at re:Invent all week long and have so much fun. See you guys next time.

---

; This article is entirely auto-generated using Amazon Bedrock.