Welcome to this next session on the latest advances in AI, data, and tech in football. A reminder to please ask questions at any time via the Q and A box to the right of your screen, we will answer as many as we can throughout the discussion and you can also share your thoughts and tweets on social media using the #ftfootball. Well, I’m just going to start by introducing my guests. We have the wonderful Vosse de Boode, Head of Football Analytics at AFC Ajax. We have Sudarshan Gopaladesikan? Perfect. He is the director of Football Intelligence at Atalanta BC. And we have Stephen Smith, the CEO and founder of Kitman Labs, based in the US. So, three people whose jobs 24/7 involve thinking about the applications of data, technology, and AI in the football and broader sports space. On that note, I’m aware that whenever we talk about AI, and that happens more and more these days, that can mean any number of things. So I wanna ask, maybe starting with Vosse, but we can move around the table. What do we mean when we talk about applications? What are we talking about here? Well, I think it differs a little bit from the more traditional models in that it, like the traditional models, are pretty good at prescribing what happened and, like, drawing conclusions and hoping to predict that in the future, that trend will be followed. Where as with AI, it’s a little bit more on what could have happened or what will happen in the near future, and it has a sort of self learning component to it. So it’s a little bit more, like, capable of doing complex analysis and also interlinked variables that we might not think are linked, but are but they actually are. So to to get, like, more complex structures and I think with game analysis, you have a really complex game to to analyze. It helps. Yeah. So. But in the end, it’s just a tool. Like, it’s another method to analyze. The question you’re after. For sure. But I guess you’re saying we’re moving away from the the sort of backward looking descriptive work and into something that gets closer to the the types of more fluid questions that performance analysts and such might want to answer. Yeah. And and, Stephen, what does what does that look like for you? So just again to reiterate, we’ve got two people here who work in the football soccer industry. You work across more different sports and different types of data. What does the use of AI in your job look like? Yeah. I think it it’s about supporting these guys. Right? We work with, you know, thousands of teams all over the globe in different sports. It’s really about helping them to develop tools that allow them to automate and create efficiency in the in the processes that they have day to day. I think the opportunity for AI in particular is exactly what Vosse said, which is it’s not descriptive. It’s not just looking back and reporting and visualizing. It’s really helping to aid decision-making. It’s about intelligence. Right? And I think there are key aspects of what these guys do every day that require a huge amount of effort, a lot of labor, a lot of manpower, a lot of time, And I think the role of AI is is really where one of the roles of AI is about creating efficiency and allowing them to process these tasks faster and then be able to spend more time unpacking that, working with coaches, helping to translate that and turn that into something that’s actually actionable and is understandable and can be leveraged and put into practice. So so so I was gonna just add to that. I think if we were to break it down into, let’s say, three key buckets, you have sort of work that can be done in the transfer market and player trading. You can have work that’s done in week to week match analysis trying to find those, small marginal gains to exploit weaknesses in the opponent, things like this. And then you have, especially, it’s an important topic given the fact that a lot of teams have very congested schedules. You have athlete recovery, athlete injury prevention and just, like, trying to keep players in the in the state of fitness. And I think AI can maybe help in in all three of these things. We already know that, a lot of work is already being done from the player trading space. And maybe now there’s gonna be seeing more things done in the match analysis and and maybe in the future on the health space as well. For sure. So on that much analysis point, people for years have said, you could call these as sort of tech skeptics, data skeptics have said that football is too fluid. There’s too much going on. It’s not just like chess, for example, where you’ve got a very sort of constrained space. So they would argue that data and AI of fundamentally gonna be limited in what they can do in football. As people who work in this space, to what extent can we now demonstrate that that was maybe they had a point or maybe they didn’t based on how things have changed over the last ten years. Like, what’s the cutting edge of these applications look like today? I think it’s always funny when we talk about sort of development. We always think, like, the point we’re at now, that’s where we’re done developing sort of thing. Like, back in the day, we saw, like, a chess computer and we thought, well, that’s possible in football. Now I would say today, we’re nowhere there where we can say, okay, AI will solve all our problems because twenty two people moving simultaneously, having different skill sets instead of, like, having chess one piece at the time and, like, lesser variety of skills. We would say now it’s not the solution to, like, sort of decoding the whole game at once. But I’m also a little bit hesitant in saying it will never we’ll never get to the point where we can actually analyze to, like, a large extent and sort of predict what would be logic, like sequences and stuff in football. So Yeah. I’m a little bit in this sort of I’m I’m a little bit conscious saying, we can never do it because I think that’s is where you’re developing, and we find something new all the time, like two years ago, no one had ever heard of OpenAI and ChatGPT, and now we’re all using it. But on the other side, we’re definitely not there yet where we say, oh, yeah. Well, we can just kinda sort of predict how this game is going based on the starting, starting eleven or something. Sure. And so what has what are the things that have changed to to improve where we are today? So, obviously, you’ve got the hardware side of things, like, you can run bigger models and that kind of thing now and and run them faster. But what about in the actual data that we’re using? Like, what data points, Stephen, maybe, you wanna come in on this? Do we have today that we didn’t say five years? I think the quality of some of the data that we’re getting now, I would say over the last twelve months, we’re really starting to see a shift. So when you think of three, four years ago, all we were getting was x y z data on player, center of mass data on players, we’re getting event data that we can cope with that. We’re now at a point that we’re getting biomechanical data. We’re getting limb tracking data. We’re understanding what’s happening at their feet, at their ankles, at their knees, at their hips, we’re understanding where they’re looking, and we can we can get both sides of the ball where as previously the detailed data was coming from wearable tech that you could see your team and nobody else’s. Now with the advances that we’re seeing in some of, like, the biomechanical and limb tracking technologies, that’s actually allowing us not just to understand the physical toll, not just the event data, but what the person is doing, how they’re looking, how they’re scanning, that opens up a whole new breath of opportunities around decision making, and really presents a huge opportunity as well. I think in terms of tactical technical analysis, and that’s that’s one area that we are incredibly excited about. We see that as kind of the next frontier in terms of really detailed analysis now because it’s not just it’s not just where was the person, it’s what were they actually doing, what were they looking at, you know, how were they acting in that? And also in terms of efficiency, because if you look at traditional assist. It’s mainly video, and still today, like, players and and and trainers do so consume video because that’s just the actual thing that they can relate to, where as if I present, like, a bar graph, that’s not something that comes to the top of mind when they’re able Don’t tell me that. On the pitch. So you have to relate it back to a situation that they are familiar with, and I think that the more complex modeling that we can do now, especially with like, overlaying event data. So that’s, like, what happened, position data, where did it happen, and then video what did it look like and then use the models to automatically select clips where we, for example, got a counter attack after a certain, fault in defense. That saves the video unless a lot of money in trying going through all the matches and trying to find those moments. You’re like automatically pop up we can do it live. So, yeah, those are things that have just made your work a lot more interesting. For one, and I think, also raises the quality because instead of three matches, you can all of a sudden choose from, like, all the matches in a season or multiple seasons. Think if I if I were to dream a little bit, we always talk about match analysis, game analysis. Okay. You know, game is ninety minutes, but the team is training between match a minus four all the way to match a minus one, and they’re spending a lot of time in training. And if you think about what a training session looks like a training session looks like a collection of various different drills that the coaches is created. And the whole idea about, you know, a coach creating all these drills is he hopes or he or she hopes that, okay, the drills that we’re gonna work on for these players, they’re most likely going to be the same sort of patterns or the patterns of play that we’re gonna see in a match. And so to just go to your chess analogy, if I were to sort of, like, dream what could be potential with AI and the advent of AI. It’s, yeah, the game is twenty two players, but if you look at the ball, And maybe the first radius or the second radius of what’s happening around the ball, there is a small sided game that’s going on, which is like a four v three or a three v three situation. And I do think AI with the generate with the ability to just find patterns or relationships between players, which you couldn’t maybe do from, like, a traditional statistical model will help us say, okay, this four v three situation happened, and then this four v three situation then turned into a five v six situation. And then this five e six situation turned into a two on one against the keeper. And then if we can talk about the game kind of like sequentially as these small side games, It actually sounds like chess. Oh, knight b one moved to square c seven. I’m not a chess expert at all, but hopefully that made sense. But I think AI then can help us sort of get to this understanding of, oh, wow. What’s the cause and effect of all these different things? And then finally, which is, like, the biggest dream is what is the ROI of all these different training drills? Like, we spend, three hours working on corners, or we spent three hours working on counter attacks. Do we actually see that ROI in terms of XG generated or goals scored? It matches. And then we can understand. It’s like even simpler. Like, how much time do we spend in each mini game on the pitch and how much time do we train? And how much better does it get? Yeah. That’s yeah. Yeah. No. It’s hard to be a dreamer. So a question on that then, and with with Stephen or anyone is, what data points would you need to be able to do that? Like, if you wanna be tracking these training sessions, like, obviously, you know, lengths of training session, great. But I presume you’d need more to know, like, what’s your when you’re throwing this into a model and seeing, did this work for a for a player or for a team? What are all the bits of information you would want on those drills to be able to make that assessment? I think what we’re we’re certainly seeing a lot of teams and coaches trying to understand what is the impact of what we’re doing, what is, you know, what are the right drills to use, like, what is actually eliciting, you know, the the outcome in terms of, like, the product of performance in a game you know, when we get a certain amount of time with our athletes each week, are we choosing the right drills? Are they actually are they having the desired impact the part of the challenge is a lot of clubs are not collecting, like, the quality type of data that are needed for that. Like, you don’t just need to understand, let’s say, the physical demands you’ve placed on an athlete in a practice session; you need to understand every drill. You need to understand what what were the coaching principles, what were the outcomes, like, tactical or technical outcomes that you’re trying to support through those drills, you also then need to understand. I think a lot a lot of times we have coaches who are trying to understand or coaching teams are trying to understand which coaches are most impactful, which ones are actually delivering it. And all of your athletes are not involved in every single drill in a training session, all of your coaches are not involved in every single drill. So that means that you need to understand who participated in every single drill, what coaches participated in every single drill. So if we wanna get smart, we actually have to have a huge breadth of context to the data as well, which in most organizations, today is just not happening. Yeah. You can use computer vision for that. That will be great. It it would absolutely it would absolutely change the game. But even if you speak to, like, some of the experts in the in that field today, whilst they’re capable now of collecting really high quality data in a game, they have the entire field map that they have all the lines when you go to a practice or or training facility for somebody. It’s a completely different thing, and they don’t have the sophistication yet. But even then, like, a lot of clubs now have access to position an event data, but that’s only like what happens where. Yep. When. Exactly. Whereas, I think our club, relies highly also on our youth academy. And then the question is may mainly, like, how did they do it? Yep. And then that’s where the next 3D data that’s now coming from video, and it’s not a hundred percent accurate. It’s not like what you’ve got in the lab, but we’re getting there. And that’s when it opens up, like, a whole new dataset. It opens up, like, actually, how were they able to perform such an amazing skill. And I think that’s gonna be, again, like, really interesting and sort of, stares our, our vision a little bit on what we should be focusing on. Because in the end, we we pay a lot of money for players that can do exceptional things. Yeah. But we hardly ever measure how acceptive, like, how exceptional it is. Yeah. And so is we’re talking about things like the body pose Yeah. Like skeletal data. So, like, how, like, where are they scanning? Like, what was their, like, do Like, people always talk about, like, one thing that makes to Brian a great is that he will be already he’s he’s it’s not just where he is. It’s how where he’s facing. Yeah. And, like, even body position, receiving a ball really sort of prescribes where they can move away quick or you’re gonna be stuck. So if you so if you got to the point where you can look at that into the under fourteens and you can be like, okay, here are the players who look really good for this already, and then who are the ones who need to focus a bit more on teaching with Martin, and that kind of thing. Cool. Stephen, a broader question for you. So you work across sports, you you’ve spent a lot of time in the US. What the the sort of simple question first is American football, real football. Contrasting the two, which which is in a better position at the moment to be taking advantage of AI in terms of the data it’s got. I think it’s different everywhere. I don’t think you can say one organization is in a better position. I think what we see in in the US is that areas like skyding and recruitment, they’re exceptional in because they don’t have, like, academy systems like we have in the European model and in real football. So they, you know, they have to be excellent at that because they’re not thinking about player developments. They’re, like, they’re thinking about seasons. Like, just that’s it. And I think what we’ve done an incredible job in Europe on is, I think, player development, player health and safety, sports science aspects like that. So I think they’re far we’re far more advanced in the European model than the US model in areas like that. And I think that’s really just, like, the culture, the way their game is designed, the way their, like, academy or lack of an academy system is set up. And I think the what what it says suggests is that there’s a huge opportunity for more knowledge sharing across, like, different sports to, like, take what’s worked and hasn’t working in different organizations. I think there’s also a huge opportunity for, like, every organization globally really to go and adapt data and analytics all the way through and the spectrum of how they operate where as I think today, most clubs are really good in one or two areas, not across everything. Yeah. You’re like good in in sort of finding an answer but then actually sort of you can know how something is done, but being able to do it is a different thing. Right? So to really apply it and to translate it into something that the players can relate to, I think that’s that’s overlooked sometimes. The translation of, like, what that actually means and how you turn, like, data science and analytics into, like, you have to use coaching language, how you use, like, football language, and how we make it accessible for people, I think that has been lost. And I think it’s it’s why in a lot of clubs, it’s not adopted well. I think that where it was most effectively applied for us when we did more complex modeling, for example, on where there’s space for, like, a good attacking pass option. It was when we actually overlaid the areas that the model showed on the pitch, like, on the video so they can see themselves running and they’re really like, at this moment, there was a better pause option. At this moment, this player was in a better position. And I mean, that’s something that you can do watching one clip, but it’s really nice if you have like the outcome of success and unsuccessful And you can really see, okay, these are fifty clips, these were the unsuccessful ones, these were the successful ones, and you can immediately use that as a sort of study material. Almost like augmented reality type thing. It’s like, sort of. Yeah. We wanna just take what we’ve already got. Don’t remove anything from it, but just add more information. Just add, like, time of really augmented. Yeah. So a great question we’ve got from a from one of the viewers, talking about different sports, someone makes the point that in NBA, we’ve seen this sort of optimization of basketball strategies in the last decade or so as more data has come in. So more three more free for more three pointers, that kind of thing. Is there a is I mean, I don’t know if I even need to frame this as a negative, but is there is there chance that football becomes sort of narrowly optimized. We see some of these sort of almost hacks coming out where someone’s realized, oh, this this just works every time. Would that is that a thing that you think is happening? Would it be a bad thing? How would that affect the sport? There’s this there’s this joke in our area. It’s really nerdy joke. And it says it’s an an self learning algorithm coming into a bar and the bartender asks what would you like to drink? And it looks around and it’s like, what’s everyone else having? So it’s really I mean, if we’re just based it on what happened before, Yeah. It it it doesn’t really lead you to any creative creativity, and I think that’s the beauty of football to always be one step ahead. And, like, use these models to your advantage, but also know how to mislead it and to come up with, like, a a better strategy. Sure. And I guess the, like, because football styles changed, like, you know, twenty years ago. It was a different game than how it is today. So the sort of the solution that solves football today might not work at all in a couple of years anyway. So I think that’s that’s probably one thing that teams could maybe use, to sort of optimize the supply and demand of the player talent pool. Typically, a lot of the times, the the same names are always being talked about just because, It’s usually a fear of missing out or a fear of, you know, making a wrong decision. Agents are pushing the same names, typically, things like this, and that then tends to sort of lot of teams tend to self limit sort of what is the talent pool that they actually go for. I think, with sort of using AI or data and having different strategies. Tactical flexibility, role, flexibility, for a player is, or a team is quite big, and you can actually then maybe improve sort of the talent pool that you look at. So instead of just constantly recycling through the same fifteen, twenty names of, like, okay. Every, you know, every competitor that’s have the same revenue profile of Atalanta’s looking at the same fifteen, twenty names. Maybe with these sort of strategies, maybe we can’t, you know, hundred percent to optimize or exploit these all the time, but we could use these at the right time at the right window to then maybe say, you know, our talent pool that we look at is not just twenty people. Gonna look at thirty five people. And here, we might be able to find some some interesting, you know, value or, things that, you know, other teams might not be looking at. Yeah. Absolutely. And and then an interesting question here about what what all this means for players, you know, we’re all here talking about in sort of applying things to people. Again, of reference to NBA and some some basket ballers being not super keen on the wearable tech side of things. All of them are monitoring. Stephen may be one for you. Like, does that come up, like, is this is this an issue we need to be concerned about? Did people broadly seem to embrace it? I think US sports are highly unionized. So there has been fear of, like, data collection because because of that because they’re concerned that will impact their, you know, their pockets, which is fair. But the reality is these organizations are investing huge amounts in in talent, and the the cost of talent is, like, soaring at an incredible rate in every sport globally. And I think that means being able to leverage data to be able to look after them better, to be able to help them to be the best that they can be and, and, like, unleash their potential that’s the opportunity. And I think that’s what any of the people that we’re talking to, and certainly what these guys are doing every day is about maximizing that potential. It’s certainly not about trying to take away from athletes. We and that we don’t have a game without them. Right? They’re, like, talent beats everything and, our jobs to try and support them and help them in any way that we can. So I certainly welcome a future where you know, they embrace and and and adopt technology and analytics like that. And it’s also why certain things like some of the optical and type, you know, technology that’s coming right now is going to help because we don’t have to ask them to put it on. We just it’s it’s collected automatically, and that’s probably why AI is is so exciting as well because it can give us more data without the reliance on human input. Yeah. And I think we’re dealing now with Generation Z that’s so much more digital native. They’re so used to, like, getting ratings, like, getting likes, getting, like, before you would go to a restaurant and ask a friend, what a good one. Now you just go online and, like, check the ratings, see if it’s something for you. And they have a completely different way of of consuming information. And I think we should be like, ready for when they come with their questions because they will be a lot bigger than what we’ve been used from previous generations. Being new to all this. Yeah. So you so you’re noticing, like, among players who were born, say, post 2000, They have a different approach way of thinking about stats and data to say someone born in the eighties. Well, at first, they’re a lot more individually thinking, like, how am I performing within the team? What is it doing for me? And they’re they’re way more used to sort of getting all this information anyway. So, yeah, I only see it developing more and more that they wanna sort of take control of their, basically, very short career. Like, it’s it’s ten, maybe if you’re lucky fifteen years. And and it’s a short time to develop. So why not grab all the information that you can and be like the best informed player out there I mean, wouldn’t harm. Yeah. Yeah. I think, Austin and I were talking about this before the session today. It’s, not just with the advent of AI, we have more information that we can give to the player, but with things like OpenAI and ChatGPT to show that, hey, we can have a concept which is kind of like I can give a without giving a name, I can give an example of, there was a striker that really wanted to sort of, you know, work on his ability to score more goals and natural. You’re a striker. But without realizing, you know, the importance of small things like link up play or without import realizing the importance of things like holding the back, you’re back to the goal and and being able to hold the ball, something that, like, Lukaku is very good at Roma. And, you know, with the at the end of AI, you can just say, hey, show me, like, all reference players that actually score a lot of goals, but also do these, like, you know, eating your vegetables. So doing all these other things that are important for you to be a top striker, can you show me these examples of these players doing it in video? Can you show me these examples of these players doing it against top opposition. And AI can then maybe sort of work towards how the player best learns and maybe can be used as like a pedagogical tool. So that we have, you know, for this audio learner, we have maybe audio. We have for this visual type learner, we have video and things like this. For someone who does like bar graphs, We have biographs and scattered parts for that. Yeah. There might be a player that likes biograph because that’s really interesting though because my ignorant question was gonna be, is there not a danger that with players being more stat savvy, they become more selfish because they’re more looking at their numbers and optimizing those. But you’re you make the very good that would now make perfect sense that it’s more that players in maybe twenty years ago were just doing that. They were more focused on scoring their goals because they didn’t have all these other stats, like, expected assists, and and just as you say, all the other footage of how a player can contribute, and now a player can actually appreciate and quantify more things that they do. Nice. Cool. And so another point that came up earlier was agents. So I’m interested. Could AI disrupt the sports agent business by sort of stripping out a bit more of the, like, politics and smokes and mirrors and who you know and what you know. And more clearly quantifying how much a player is worth. Do you think there are either agents out there thinking what happens when my player can just go in with his own spreadsheet to negotiate? Or, b, do you think there are other agents who are now bringing more and more data into their agency with the knowledge that this is gonna become more important in negotiations. I think it just comes back to, like, being really well informed. And I think being an agent is, like, way more than that. So I would only use it in my benefit if I was one. I don’t think it it’s gonna replace any of that. I mean, it might give, like, clear, like, even a better overview of, like, hey, this Club is maybe not so keen, but there’s, like, three other Clubs that fit your profile rights. That’s try to. So I think it just opens opportunity rather than closes doors. Sure. Sure. And, well, similar similar question I was just gonna ask for you three. How is what’s the potential of AI to disrupt your jobs? And I don’t I don’t, to be clear, when I say disrupt, I don’t mean replace, but just how will your jobs look different or how do they today versus ten years ago, five years ago, have you had to have different skills? Does does the day to day look different based on to this new technology? I yeah. I think that, like, the the most important part of my job is to use analytics, but then translate it. So I don’t care if it’s an AI model or just like a logistic regression that we’ve been using for ages or just like summing up the number of counter attacks versus the number of corners that we score. To me, it’s like, I have to make sure that it fits in the goals that we’re working on, and it fits like the team focus or the Club’s strategy. So it starts with strategy, and then answering questions that helps you to, like, make that strategy better to reach your goals. And to me, like AI is just another great tool. And I hope that there will be plenty other great tools over the next so many years. So the core of the job to me is the same. It’s just that the methods change a little bit. Sure. I think, yeah, I think for me, we’re already seeing a lot of competitive advantages that teams are using with data in player trading and match analysis and player development things like this. And that’s gonna continue, I think, to improve, to a point where maybe a lot of the stuff is automated from the tactics side. There’s always gonna be things that you can improve and things that you can discover new. But it might give us a little bit more space to think about things that are harder to solve: How do we make the best of what we have right now? How do we make the best out of the squad that we have right now, not from a tactical point of view, but from a player motivation point of view, from getting two players to to work together better, from character development. I think one of the biggest things that we think about, it’s, you know, This player is is rated quite high on some models, but does the player have the right characteristics from, like, a behavior personality perspective. And maybe we can then start working a little bit more on squad construction, not just from the tactical technical side, but squad construction in terms of what does it mean to have the right number senior players, the right number of young players coming through. What does it mean to have the right number of mentors, and really work on sort of those, like, softer elements of squad construction, and then just let Chelsea know when you get the answers. But but just talk me through on that one, like, how does data come into that? How does tent data technology AI come into that because that’s I completely agree with you, but what’s Oh, I was just saying I was just saying that data data helps sort of give us, it can’t data can’t solve all tactical things yet, but data gives us a nice plateau where we can feel comfortable with, okay, we’re able to evaluate a player’s technical tactical ability. Which then gives us as human beings the ability to use our intuition, our empathetic sense, our emotional intelligence, to then focus on problems that maybe we can’t use data yet for. That’s how it’s shifting the landscape in my in my realm completely because you can hear it from both of these have said is that it’s no longer we’re no longer fixated just on solving just the tactical element, just the technical element. Actually, the the breath and comprehensive nature of the questions that they want to answer are huge. And that means that there’s a complete shift in movement from having fragmented data to having very, like, highly rich contextual comprehensive data that you can extract performance intelligence from. It’s not just about solving, like, a workflow issue and doing something over here, it’s actually about answering really interesting, like, really, like, impactful questions. And that means that the shift for people in the tech space, like me, is that, like, the, like, the complexity of the problems that we are solving has like, multiply it by a factor. And that, you know, that creates a huge amount of challenges. And I think it means that if you’re a tech company, you’re you’re in this space and you’re focused on one small myopic sliver, you’re in trouble. And if you’re focused on, you know, being a being a database, I think it’s over for you already because the last thing this isn’t this industry needs is another data. Right? We need intelligence. We need insights. We need things that we can action. So yeah, it’s almost like the classic example of what people often say about AI is it doesn’t get rid of jobs. It gets rid of tasks and gives you more space to do bigger and better tasks. And it it feels like in sports, this is more applicable almost than anywhere else. Well, I think one of the things if you think about the larger language models like Open AI, they could sort of open up analytics to a broader audience because it sort of lowers the level of expertise that you need to ask questions to datasets. So that will happen on a large scale as well. On the other hand, I can also already see it where there’s, like, quite sort of not very thought through models or, like, not very tested well models, and people start using them as a truth because it’s a number, so it must be right, without taking to account that you’re evaluating player without taking the competition he’s coming from into account or that your study played with, and you’re not aware that the model doesn’t take this into account, but you put this out into, like, a flashy dashboard and that’s your new truth. I feel like that’s, like, one of the mistakes that some Clubs are now making, which is the new director thinking, oh, I got this great tool. Has a lot of intelligence behind it. It has AI, so it must be great. And then this is my truth where in the end, if you are decision-making or in the club, you have the responsibility to to at least to some extent, understand what it’s doing and what it’s not doing because if you make mistakes based on it, you are responsible, you can’t put the algorithm behind bars because it’s racist, right? You it’s it’s it’s you who were responsible for developing that model. Yeah. And and the same is, like, if you’re gonna spend twenty million on a player, you you really well have to know what the model can and cannot do Yeah. It might look flashy if it’s new, but they’re, like, don’t be afraid to ask questions and say, how does it actually so we were talking about this before we came in here about this this idea that in football, you’re almost you’re in terms of an evidence based role, you’re so close to the outcomes. You’re so close to the evidence that, like, there’s nowhere to hide. You can’t you’re not working on a product for five years and finding out later. It’s you’re finding out on Saturday. Like, the the downside risk is like you’ve just said. Like, you know, if it’s wrong, you’re there’s almost a risk of the baby being thrown out with the bathwater, like, the whole strategy that you’ve been working on someone says, you told us to do this. You told to sign this player or you told us to take set pieces in this way. How do you manage that risk? The the classic, like, with great power great responsibility thing. Like, what conversations are you having? How you manage your expectations about what your your tools can and can’t do? Yeah. I agree with you completely. I think the biggest the biggest sort of mistake that could happen within the team is is one, breaking your wage structure or two, spending a lot on your transfer fee and that could maybe affect your next three to four windows, especially in player trading. I wanna maybe do, like, a small example from Benfica times when we were trying to understand how do we use data to evaluate, young players coming through in which players maybe might be right for Benfica’s Academy. You’ll be surprised by just how much you can do just by using the metric of the number of minutes someone’s playing. And, obviously, you know, a thirteen year old kid who’s playing tons of minutes in U thirteen level, but he’s also getting non trivial minutes at the U fifteen level in this local region in Portugal. It’s probably a good sign that, you know, there’s some talent there. Maybe we don’t know all the exact characteristics that the care that the player has, but he’s a thirteen year old playing a lot. On Saturday, and then he’s playing a lot on Friday for the U fifteen’s. It’s quite interesting, you know, data point. And I think for us, it’s I think for me, what I would say is there is a lot that you can do right now, with the data that we have. Even if you just have event data, and not this tracking x y data, you can still create good competitive advantages. If you know what your vision is, what questions, you wanna ask what your objective is if you’re a team that wants to be winning games, if you’re a team that wants to increase perhaps, you know, the the squad value of your academy players are being really clear on what your objective is. And then having sort of these, like, executive level dashboards and these statistical models that maybe use careful relationships between, dribbling, passing, defending blah blah blah, and and then winning. And then once you have this sort of, like, baseline set and you say you know, even if our AI strategy fails or our AI strategy is gonna tell us something like shoot closer to the goal, it’s probably gonna be better for you to score. Something very obvious, you know. Like, it’s okay. You you you can always fall back on this this approach that works because there are teams that are being successful today with an approach without necessarily using AI. Yeah. And then once you have this sort of, like, common baseline, go full force. So picking up on that point about some teams using AI, some teams not. A question from from one of the viewers here saying, do you see performance led AI becoming accessible at every level of the game. Is this something that will be, like, democratized, or is this something that will accrue to, like, the biggest clubs and organizations already? I think, like, Open AI is now open to everyone. Yeah. I think analytics will be a common thing for everyone I actually think people will enjoy looking at a game through the eyes of a player, like, pointing out where open spaces are that they would search for other than just watching a ball, which is like no offense, but a lot of the fans are still just only ball watchers. I think these models can really help to to make the game more fun to watch and and look at it from, like, literally a different perspective. I guess already, like, on Monday night football in the UK, for example, we get the the more in-depth looks at tactics and people talking about expected goals and things. And it’s just you’re talking about just a an acceleration or continuation of that trend with more and more, like, thorough understanding of what’s going on on the game. There’s an opportunity I think to create, like, an intimacy within the game for fans as well. I think the more advanced AI and analytics become, I think, the more explainable, essentially, like, what outstanding performance looks like is. I think to a large extent, we’ll be able to decode performance. And I think if we can make it explainable enough, that means that I think broadcast and media will be able to start to leverage that and really understand, like, what makes people incredible, what makes certain things incredible. What’s this? Yeah. Exactly. Also, I have a friend who just started a job at an initiative around women’s football, Mercury Thirteen. And we speak about this a lot in terms of the women’s game is is probably different than the men’s game. I’ve had maybe a very short hat, like, less than a year experience working with the women’s team in the US, and they seem much more interested in communicating with one another and figuring out how do we advance the ball closer to the goal as, like, you know, groups of three, groups of four. And then sometimes you might have a team on the men’s side where it’s like, okay, I have the pace and power. Or if you just get me the ball, I can get to the goal myself, which is totally okay. There’s there’s there’s multiple ways to to get, you know, the ball closer to the goal. And I think maybe AI can help us sort of appreciate and celebrate the fact that football is played differently. And we can tell those stories better of you know, just to go back to, like, the fan engagement piece, because, yeah, women’s football and men’s football apply two different two different types of football. Yeah. Yeah. And then, one one last last question as as we’re coming short on time is like, do you see the other roles like senior football roles, like whether whichever football we’re talking about, coaches, managers, are they becoming more comfortable talking about these things, are they becoming almost more quantitative? Or is it more that that’s all going on sort of over here in a separate room and the managers are still the managers they’ve always been? Several instruments. I certainly think they’re becoming more inquisitive. No. That’s the the word I would I would probably use. I think most sports we’re seeing, there’s a move towards leveraging data and analytics. And I think that means there’s an appetite. And I think, you know, I had this conversation with somebody a couple days ago at at a conference last week, and they were saying we’re seeing a lot of younger coaches come in and that we think that’s changing it. And I actually I think yes, it is, but also there’s a lot of very experienced coaches that we work with under that the game is evolving and that, like, for them to, like, stay doing what they’re doing every day, they need to evolve as well. And most cultures are pretty smart. But if you ask him, would you be like the best informed coach out there or the least informed? What would they what would they choose? Yeah. That’s basically what you’re doing. Right? And they can make the decision, but at least they are informed. Yep. Yeah. Well, folks, thank you so much for this. Boss has said, Stephen, this has been fantastic conversation. Really appreciated it. So I just wanna thank everyone else for watching as well and for these great questions that got contributed. And a reminder, this session will be available on demand for ninety days on the event website. In a few minutes, my colleague, James Fontanella Khan, the FT’s US corporate finance and deals editor will be chairing a panel on US soccer with Jessica O’Neil, president of Business Operations at Houston Dynamo, Steve Horowitz, partner at inner circle sports, and Cara Nordman, cofounder of Angel City SC. We’ll be back shortly.