49: How AI is changing sports

In high school I did a lot of stuff for the football team, including taking all the video or back then it was films and we would have to watch the films with the football team, try to learn from them. But now we're in an AI age and we're about to change all that in a deep way. So that's what we're going to talk about right now. Who are you?

Hi everyone, I'm Max, co founder and CEO of SCORE Technologies. And as you just said, we, we're using computer vision to understand the world of sports.

Yeah, well, what does that mean? Yeah, you know, do you. Well, let's go up to 30,000ft. What kind of sports? You know, football is an obvious one because you can do a lot of analysis and help the teams out and really understand how various teams, you know, strategize their plays and, and come up with their place. Right, but what is it all you. That you all are doing?

Yeah, so obviously when I'm what, I'm what I'm saying sports. So funny enough, we're doing football and soccer. So I don't know if, I don't know if you're referring to soccer or American football, but we're doing, we're doing both and we're doing other sports. But we really started with soccer because we are just, you know, from our backgrounds, the co founders, we are really connected to this world. Understanding sports with AI is, is a very interesting topic. I would say from a high level perspective when you watch sports you have different types of data. First one is tracking, so you understand the movement in a very simple way. Second type of data that you have is events. So something's happening on the field, there's a foul or you know, whatever, a touchdown, a goal, whatever third one is to apply a grade a score to, to what's happening. So literally trying to understand the impact of what happens on the outcome or on a goal, on a touchdown, yada yada. So yeah, I could, I could go deeper, I could explain from a tech perspective. But high level, we just want machines to understand the impact of movement. And the reason why sports is very interesting to do that is because it's really complex. So it's moving, people are colliding, are doing a lot of different, very dynamic movements. And also it looks good. I mean when you just talk about this, everyone gets interested. So that was a great way for us to go to market. Definitely.

And is this for the teams themselves or is this going to be a new kind of stats on the super bowl or the World cup that we're going to see on TV someday. We're going to see all sorts of new stats, but you know, how, how fast is the receiver running down the field? How much faster is he than the, than the defensive lineman who's trying to chase him?

So hopefully you'll see, you see like powered by score, you know, during the super bowl pretty soon. But we're not working on this at the moment. What we're doing is we're trying to bring this level of AI to clubs, you know, and scouts. And the reason we're using, and maybe we'll have a conversation on this, but we're using collective intelligence. We're using decentralized AI to do what we do. And the reason we do this is because it allows us to have access to the same level of intelligence even if we are provided with low quality video. So you have a lot of competitors that are focusing on the elite, which is good, and we can do it as well. But our vision is we have to go kind of go deeper and try to get access to lots of different videos. So it creates new challenges. But yeah, going back to your point, we're trying to help scouts and clubs to kind of get access to a nice computer vision framework.

Yeah. And, and is this something that a high school team can afford or is this only for the pro, the pro sports?

Well, that's the whole point. We have a, we have a strong thesis at school. We, we really believe that, you know, big teams and franchises and, and anyone that has deep, deep pockets wouldn't try to get access to that kind of open source technology. So the reason we're doing this with collective intelligence, the reason we're doing this open source is because we think the whole world needs to have access to good computer vision. And obviously, yeah, so obviously high schools and even grassroots. There's a funny thing we share internally is that all the parents that are watching their kids play on Sundays, they're all recording videos. So obviously one of our dreams is to be able to kind of quantify this and create a framework that, you know, will allow older system to kind of see if, you know, someone is good enough to kind of go to a different level and all. So yeah, obviously the dream is to bring this to all the different levels.

What kinds of things can attract, you know, I assume you're looking at formations and I assume you're looking at speed of runner. You, you know, completion rates, stuff like that. You know, what kinds of things does your system track? And, and I, I didn't Ask. So do you have a demo ready? Do you have a demo that you can show us what it sort of does on the field or so.

So I have a demo that I can share with you after, after this call. And I don't know if you would be able to add it to the podcast.

Yeah, I will.

Yeah, I'll send you something. It's. It's in soccer, though, so you'll have some, Some. Some footage.

I like soccer. I saw Pale do a back.

Back.

He came in and visited me. Because you had to be a certain age to see Filet play.

Yeah. So the sort of data we're getting is pretty much what you described. So when I, when I was talking about tracking, the first thing we do with our framework is we annotate everything that's happening on the screen. So literally any movement frame by frame, so we know everything that happens on the, on the pitch. And then the second step is to kind of run neural networks, so deep learning algorithms, so that machines can tell you, oh, that was a, that was a foul, or, you know, that was an offside. You know, offside from the offense or the defense if it's American football, or just an offside if it's soccer. So, yeah, so different level of understanding. Some customers just need the annotation part, so they just need to understand the simple movements. Some customers would need the whole, I would say, workflow from what happened on the screen to kind of a plain English insight telling them something happened. Or you should have done this differently or maybe this player shouldn't play in this formation. Yada, yada, yada.

I met with an eye doctor down in South Africa who works with professional athletes, and she told me about one baseball player, actually that couldn't catch a ball coming at a certain angle. His eyes and perception system had a slight flaw there, and she, she fixed his eyes with her program. Is this the kind of thing that you can identify about a single player, that they have a weakness at a specific thing, like catching a ball at a certain angle or seeing a formation that was, you know, in front of them? They keep making the same mistake that maybe they need to get.

You know.

Some work with her. You know, that's tune up his brain.

Yeah, that's a very, very good example, because what we realized, and I mean, we were not the first to realize this, but obviously a team is made out of different players. And so the way we are building our scoring system, we call it the player value function, is that we're trying to understand at any given moment in the, in the game, what's the impact of a player on, on an outcome, but also on the performance on the, of the player next to them. And so I, I think for sports like football or soccer, we wouldn't be able to kind of tell about the angle of the ball probably if we just focus on baseball more or cricket, that would probably be something we would be able to do. But yeah, that's the whole point. The whole point is to try to find inefficiencies or way, you know, room for improvements. That's, that's the whole point. And, and that's also the reason why our company is called Score. People often, oftentimes think it's because we are just a sports company, but the reality is because we're building a scoring system.

Wow. So, yeah, each player is going to have a little bunch of scores. You know, I, I've been reading some scouting reports for American football, and they really go into depth about the.

What'S.

Their demeanor on the field? You know, not just can they run, you know, 50 yards real fast, but they're looking at the holistic person because that first, that affects their value in, in the professional leagues. Is this something that your, your system can help, you know, really try?

Yeah. And, and so we're connected to a lot of athletes, some of them being top, top level. And I think we, we, we don't really like, you know, these counting reports. I mean, I think we can all think about Tom Brady's scouting report where everyone was telling, like, this guy would never do a thing in NFL and it's just like the best player ever. So I think the reason we went with this approach, trying to be open source, right. To be unbiased is because I think we need that. We need a very, you know, unbiased and objective way of scoring players. And it will bring, I think, new insights and also opportunities for people that are not really fitting in the current system, which is based on, as you just said, on like one individual or maybe a group of individuals are trying to judge and assess the performance of a given player. So, yeah, I think AI in this case could probably be positive. Yeah.

One big problem with sports is injuries. I mean, injuries can dramatically change what you can earn as a player. Is your system looking for behaviors that could lead to injury and sort of telling the player, hey, you're hitting this person with your head down. That can lead to a concussion or a spinal injury or something like that? Why don't we change, you know, work with the coaches to change how you're you're hitting that person so that you have a better chance of having a long career.

So, yeah, that's. That's a fair point. That's probably one of the first conversation we always have with players when we talk about what we're trying to achieve, they're talking about injuries. Just to give you kind of an overall idea of how the scoring system works. We have a background in AI and finance mostly, and we're trying to have the same approach. People are talking a lot about quants and stuff like that, but that's pretty much what we do. So what we want to do is we want to collect all the available data, and then we want to find patterns within this data. So if we need to navigate through the data to really focus on injuries, we would get an insight from this data. And our scoring system is made out of all of different data points. But to talk specifically on injuries, what we're trying to understand, and it's not done yet, we don't have a full ready product for that. But what we're trying to understand is, as you said, some correlations. Do players have a higher chance of being injured if they do false prints in, you know, in three, four, five different actions or something like that? So it's very interesting to try to find correlations. And sometimes you find something that is completely counterintuitive. And I think that's when really professionals can really get on your data and understand that actually they could do something different.

Yeah, it, you know, every player is different and has different strengths and weaknesses. Right. One guy might be really fast at running 50 yards. Another guy might be slower, but be able to spin real fast or fool a defender and get extra yardage that way. Does it. How does the score look at those kinds of things? And does it come up with a single number or does it come up, this guy's really fast, but he's really, you know, or this guy's really slow, but he's really slippery. You know, he always gets extra yards because he's able to fool people, right?

Yeah, that's. That's super interesting. And, and hopefully you'll be able to show on the screen what. How we're doing it. But it's like a heat map. And on the heat map, you have the position of the players, but also you have the position of the ball or you have the formations and a heat map. And frame by frame, so you can literally go through the whole game. You would understand who was kind of bringing a different impact on the outcome. So it has to be visual. Obviously we can't just come with a large file with just numbers. So, yeah, it's a heat map. And it's very interesting because the heat map is showing you in real time what is actually the way to go. And most of the time, because we watch sports since the early days, or we're just fans or we think a team should play a way or another, and seeing this in front of your eyes is like, oh, interesting. Well, I didn't think they should be, you know, playing in this formation, but actually it should, it should be working pretty well.

Yeah. Next year we're about to get glasses that have a display inside of them that lets us see sports in 3D. Now see the football pitch on, on our coffee table in front of us, or on the floor or in the middle of the air. Using that kind of technology, I want to see a lot more stats. Are you thinking about working with the broadcast industry to start putting much, many more stats on top of the players as they're playing so that I can enjoy the game in a new way?

Yeah, completely. That's the whole part of our thesis, is to be ready for this 3D world. I think all stats at the moment, except for big companies like Amazon or Microsoft, they're doing this for big leagues and big teams. But yeah, I think this really, well, as you described it, is going to be the way we consume content. And the reason we're using collective intelligence and Bittensor is because we realized that that would be the only way for us to have access to enough compute to be able to process data faster than broadcast. So broadcast most of the time it's like 30 frames per second on our subnet. On our Bittensor subnet, we are processing things at 40ish frames per second. So we would go faster than the broadcast and then we would be able to layer the information on top of the broadcast. So, yeah, it's very important for us to be ready for this. And also we are. So when I was talking about trying to understand what's happening, I went fast. But you have to go through what we call in AI and data science simulations. So you have to understand the space. We started in 2D and now we are processing it in 3D so that we're ready to go through this phase where we would be compatible with different devices.

That's awesome. You're talking my language. A lot of my friends and family play virtual football. You know, they're putting together together teams of different players. Could you analyze that and sort of say, oh, Your team's probably going to do pretty well next season.

Yeah, fantasy leagues are. Yeah, yeah. We have a lot of inbounds around this. I think people realize that they could probably be performing pretty well if they were using our tech. So we see ourselves as a layer, right. We are calling ourselves like the optic nerve of AI. It's a bit fancy, but it's funny because we just want to provide this open source layer to people. So if anyone who's a big fan of fantasy football, you know, if they want to come and build something with us, we'd be more than happy. We would probably help them to build the product.

To be honest, it sounds like a good thing to do because there's a lot of people out there who are willing to spend money, you know, building their fantasy football team, you know, because there's money on the line in a lot of these cases.

That's. That's the real business.

It was certainly pride, you know, hey, you got a good team this year. Oh, man. Wait, tell me about the company that's building this. What kinds of people are working there and how did you guys get funded to. To do this? Because it's not cheap to do, you know, to hire AI people, you know.

Yeah, definitely.

I hear it meta. They're paying 10 million for a good, you know, AI person. It's like crazy.

Well, that's the beauty of Bittensor. That's there's a reason we went there is because we had access to brilliant minds. And also the system is made that, you know, the incentives are not coming from us, but from the system itself. But yeah, our team is composed of different people. So as I said, I have a background in collective intelligence and AI and finance. One of my other co founders, Nigel, is connected to the world of sports for decades and also entertainment. He actually built a company with our cto Tim, also connected in the same world. We've been working together and building tech products. We actually built some products for top athletes like Harry Kang, the England captain. I don't know if you know about him. And then we realized that we should merge all of our experiences and build this kind of new way of processing data in sports. And we also have people within our team that are more, I would say, extreme in quantitative analysis or in community building and stuff like that, because we also identified pretty early that using collective intelligence was a positive but also a challenge because you have to get attention from people and you have to kind of express yourself in a good way. But yeah, as you said, it's very challenging because the competitors are hiring people with way more incentives, I would say, from an internal perspective and then from a funding perspective, so far we've bootstrapped the business and I think we're waiting for us to be ready to compete with big, big names to kind of see if raising money would. Would be interesting for us.

Yeah, it's. It's cool. Since you're studying sport at such a technological angle, what else are. What are you seeing happening in sports that's because of technology? Like what you're bringing? How is it changing the game?

Yeah, it changed the game. If, if we talk specifically about soccer, I think American sports are a bit different. I think. A bit different. Sorry. I think they were using data since probably like the Moneyball moment, you know, in. In baseball. Yeah. So I think they're pretty much advanced. What if you go to Europe and talk to soccer teams? I think two of them are really, really advanced. One of them is Brighton in the uk, and the other one is also in the UK called Brentford, and they're using the Moneyball approach as well. And the funny thing is they. They performed really, really well and they managed to build large and decent clubs. So now all the different clubs like Liverpool or other big clubs like Juventus in Italy, they're trying to kind of build this whole AI department. So what I see from the ground is a lot of appetite to kind of bring data on board. But also they're facing the same challenges, as you said, you know, AI talents in AI are hard to hire and GPUs are costly, even if you're a big team in sports. So they're facing the same problems, you know, as other companies.

Very cool. Well, thanks for giving me a little insight into what's going on with, with technology, AI and sports, because it's a fascinating field for a lot of people. Anything else that I should be asking? Because I, you know, I. I'm not a professional athlete, you know.

Yeah, no. So, no, I think we covered a lot of. A lot of parts about what we're doing. Maybe you have more questions about potential or not. That could be interesting or if you want to talk more about sports, I have to say that the biggest type of customers in this field for now are betting companies, but I'm not sure that's the type of content you want to. You want to talk about.

No, no, there's money and gambling, like I said, and they certainly want to know, you know, is there a technological way to get insight about the teams that other people don't see?

Right.

It Changes the odds of the game and stuff like that.

That's critical for them. So having access to real time data also to a different way of kind of finding the inefficiencies, it's in the end for them is, is money on the line? Because as you said, odds are drifting. They're trying to protect obviously the, the house. So it's critical for them. So that's why they're massively investing in, in that type of technologies. And just for you to know, the two clubs I've mentioned in the UK, Brentford and Brighton, are led by CEOs that are from this kind of gambling and gaming background. So there's a thin line between working for clubs or working for sportsbooks. But the reality is without the sportsbooks, I think the Moneyball approach wouldn't be as advanced as it is because money is poured into data and AI teams by the gambling world and weirdly, it's bringing a lot of innovation.

Yeah, it's funny. Money drives innovation, hires engineers. Well, thank you so much. This was really interesting. You know, AI is changing everything and with with the holodeck coming, the glasses and the new light field displays, sport is going to change to be a more 3D thing in the future. And having more stats is a big deal. You know, I've always wanted that because I knew a lot of the teams had that kind of stat. I mean, I had dinner with the guy ran South African cycling team and he knew how many falls per race each rider had. Right. Because they keep stats on all sorts of stuff like that. And that's how I found that eye doctor down in I innovator Shirley Calder, who runs IGEM down in South Africa, who works with professional athletes. So there's this whole industry of technology and brain, human performance people and gambling people over on the other side.

That's really fascinating.

Yeah, yeah. Where do we learn more about it?

Mainly on X. If you want to follow us, you can go on, you know, x.comweBuildScore I mean we're posting everything over there and we'll show you everything we do and you'll find also a demo about our product.

Yeah, very cool. Well, thanks for spending a few minutes with me, showing me what's going on with computer vision and AI and sports because it's fascinating a lot of people.

Thanks Robert. Thanks.