Is AI running your football club?

9:55Ryan Beal:And when we think of, well, how do we process that information? How do we analyze that? How do we interpret that? How do we turn that into a decision? AI holds a huge amount of promise for us in those areas, just from a pure efficiency perspective. And our ability then to be able to process that and make a decision really, really quickly, that that's what matters in an industry like this and ai holds the cards there so steven just on

10:18Gabby:that we'll come to the recruitment side of things in a moment but on the on the injury side of things and the medical side of things for all that ai provides it would still need and i apologize if these are basic questions because obviously i would you know me and ai are a long way apart at the moment. But if you have all of this data and all of this access to the information, does it still need the head of performance or the head doctor or the physio or the head of sports science to interpret that for the head coach?

10:51Ryan Beal:Yeah, one of the key tenets for us is the fact of augmentation, right? The idea of AI is not to replace a human, it's to augment their capabilities, to allow them to be able to make those decisions faster and to make them more refined and to give them more weight behind that decision through the use of analytics so the idea is never that we're going to replace that head of performance or we're going to replace that head of medicine a human in the loop is so so crucially important what we're going to do is we're going to give them better information we're going to give it to them faster and we're going to let them actually then deploy their their expertise quicker generally what's happening is that we're we're managing the athlete every single day in practice and then we're trying to get to we're trying to get to the game so the decisions are happening around how do we individualize their training every day so what should they do or shouldn't they do when they go into the into the gym what aspects of when they're walking out onto the training field because of what's going on maybe they're not doing some of the contact elements maybe they're not so doing some of the higher intensity components because they provide higher amounts of risk today i'm also

11:53Gabby:wondering steven whether it whether it simplifies everything for some people and by that i mean so i I interviewed former head of sports science at Celtic I was at Tottenham with Ange Postacoglu as well and by the end of that interview I've said this before by the end of that interview my head was absolutely scrambled with what is on his plate when it came to trying to help with planning the training sessions who needed what load who might have been dropped from Thursday but they were going to play on Sunday so how were they going to train on the Friday and the Saturday and so actually trying to plan training sessions for a premier league first team also have european football 10 days in advance was mind-boggling because of everything that has to be fed into it for each individual before you even get to what the team needs are yeah 30 individual periodized

12:43Ryan Beal:plans that are on completely different tracks because the demands for each of the players are going to be different what they've been doing is different and then what's required of them because the way they're going to be rotating is different and then you then you have to take into how the team wants to play, what the coach actually wants as well. So the level of detail and sophistication that's needed to operate at that kind of level, especially for not just a standard team, but a team, like you said, that's playing European football as well as domestic football. It is mind-blowing.

13:11Stephen Smith:Ryan, you focus a lot on recruitment. And so I imagine some of this information in terms of injuries and how long plays, all of that goes into the model. And while the head of the kind of medical department of a football club might not lose their job because they're using this information in terms of recruitment the people that are giving that information currently to the head of recruitment their jobs presumably are at risk aren't they because you're doing their jobs for them yeah i think it comes back

13:38Mark:something steven said where it's all about combining the best humans with the best ai to come up with something that is overall better for everyone um so you need the human to really trust what you're giving them and it's not always best especially in football to build the most complex model it's best to build the most explainable model and something that you can really explain

13:55Stephen Smith:why it's come to certain decisions so where has it worked well so i think i think what anybody can

14:01Mark:do well is look at any league and see who's scoring the most goals and who's the best defender in that league what's really challenging is sort of then looking at all of the all of that data around that player to look at if that player is going to then continue to perform well at a new club so it's that simulation side where i find our research has been really interesting where you can give a decision maker this extra bit of information where it's saying that we've simulated this player from bundesliga into the premier league and we can see that actually they're playing against quicker defenders they're playing against stronger defenders they're going to score 20 % less goals is that florian vertz neither i think the ai actually likes it well i don't like works it's werner who really liked a few years ago and obviously it was proved wrong but okay so let's

14:41Gabby:take timo werner then rather than florian vertz or you know several strikers who've lit up their a divizy and then come over here and and haven't managed to do it your models will simulate say Timo Werner his first season over here would have simulated what it would have expected him to

15:01Mark:do yeah so you're looking at if that player's over performing in there currently you're looking at that type of data to see are they just having an incredible season and and gochera's arsenal you look at the goals he scored in portugal and then you're saying well actually what's the the lowest level teams he's playing at and how do we compare them to someone in England so you look at the lowest teams in Portugal are actually more comparable to the lowest teams in the championship in England so you're then trying to look at what percentage of the goals have come from that quality of teams and then translating that across then you've got all the other things like tactics that change you've got the language change the teammates change so he's going to get the same quality of chances all of these type of things are so much more information that an individual scout can sort of process themselves so we help them sort of process all of it and provide that simulation

15:47Gabby:does it then and i will turn it to verts now because quite frankly i can't remember who else chelsea signed around verna to to use this example will your model then simulate how verts will do if isak is signed how verts would do if nunes was kept how verts would do if he replaced gacpo on the left how Virts would do if he was in a three with Sir Bosley and McAllister would it do all of those and then you hand them to a head of recruitment yes we break it down into looking

16:23Mark:at how your work of the team makes and how he works in a certain tactical system so I spent a lot of my PhD a few years ago looking at how players work together and what are the qualities that make a good pair of players I think a lot of data has been spent looking at how do we find individuals but actually when you come into football compared to sports like baseball where money ball is really big that way that they work together is so important like like give us an

16:45Stephen Smith:example of your PhD the pairings that you thought worked well yeah so if you look at a player like

16:50Mark:James will Prowse who is big on crosses known for set pieces and then if you paired him up with someone like Chris Wood who's really well known for headers and winning lots of set pieces you can sort of see in footballing terms why those two players might link up well together so you're going through millions and billions of pairs of players that could be created across so many leagues to then find what are those sort of qualities that create a really interesting pair of players and which pairs of players are at the central core of any team.

17:16Gabby:Although the only thing that I would say with that, without disputing that model, is I could put James Ward-Prowse and Chris Wood to me.

17:23Mark:Yeah, we're looking at sort of hundreds of different features here. I'm trying to give a sort of simple example.

17:29Stephen Smith:He did his PhD in Southampton, so I think James Ward-Prowse was obviously front and centre of your thoughts.

17:35Mark:That's probably it, yeah. Definitely could do it at the moment.

17:37Gabby:But presumably it's more nuanced in, if you put James Will Prowse and Thomas Suchek together, now does that have the legs for a West Ham midfield? So is it doing stuff that my eyes can't see?

17:55Mark:Yeah, so it's trained on millions of data points that have come before that to look at how Will Prowse and Suchek have linked up with similar players to that person in the past, and then when they connect, what happens. So obviously this is then sort of linked to who they're playing for at the time. So you've got to then feed in that other information to sort of normalise for Burnley versus Arsenal. But it's looking to see that when those two pairs of players link up, what's going to happen? Are they going to move the ball forward? Are they going to lose possession? And using all of these things over a number of different years to provide those simulations going forward.

18:26Gabby:Sorry, so sample size is the key here, isn't it? I mean, presumably you could go back through James Ward-Prowse's career all the way to youth team, could you? I mean, if that evidence was available back then and all the data was back in back then. And so sample size is the massive advantage here because you can run all sorts of different models, different games, different players, and get it into a one page for a head of recruitment quite simply.

18:54Mark:Yeah, exactly that. And it's trying to break it down into different sort of things that a head of recruitment can understand and process sort of as an independent bit of advice. So if we're looking at what's his tactical fit to a new team, we can simulate that forward and give that as a, that this player is used to playing in this formation, used to playing in these tactics. This is what your manager plays and this is the risk that you're taking in terms of how that player can adapt. So it becomes a sort of bow in the arm. Yeah, sort of a bow in the quiver sort of thing. Yeah, yeah.

19:20Stephen Smith:Across the summer's transfers, as far as you know, how much of the decision-making across the Premier League was using AI?

19:27Mark:yeah I think it's still coming into the game more and more I think if we look at Brighton and Brentford and what they've done over the recent years with their owners both coming from a betting background where they've got some of the most complex data in the world because they've been motivated in their other jobs which has created their fortunes to be able to buy their clubs that data is then processed in a different way by them and clearly it's worked on the pitch and even now I think Brighton have spun out their analytics platform into Jamestown's analytics which is driving hearts and we can now see it's working at hearts as well in scotland so you can really start to see the benefit of those clubs and how they're using data better than anyone else you've then got teams like liverpool and man city who have gone down the more physicist approach they've hired harvard professors and physicists to process tracking data so they've got all of these players on the pitch that they treat as sort of tracking particle physics and they've built all of these incredible models over the top which are using some ai to learn so they're using the same

20:21Stephen Smith:systems?

20:22Mark:They've sort of taken similar approaches and sort of hired similar people to do that. I think Liverpool were the first to do it and then Man City have sort of followed suit. But yeah that's then informing more the match analytics side compared to the recruitment side like Brighton and Brentford. And then even now I think you saw yesterday Reading. And does that feed into tactics?

20:39Stephen Smith:Yes definitely. So that information will then be spread to Pep Guardiola? Yeah.

20:43Mark:So I think that translation layer is really important. It needs somebody to understand what's coming out of that being fed into the coaches, being fed into Pep who can then sort of execute on this information and sometimes you might ignore it i'm sure there's there's a case of yeah translation and that trust of the models but you've got the best people in the world building those models so there's there's that element of trust that yeah a club like city we've looked at the recruitment

21:04Gabby:we've looked at we've looked at the medical side of things gabby touched on tactics there and we heard from laura harvey at the start of the pod who's who's used it in the american women's lead for tactics will matches be increasingly simulated through ai by coaches before an actual match to go through various tactical nuances some of that's already happening so i think they are

21:29Ryan Beal:looking at that and even some of the the the examples that we've you know just discussed i think those simulations are happening with different um makeups with different tactics with different strategies in place with different combinations of players and i think the the more, I don't want to say the word advanced, but some of the more sophisticated teams are using those tools to, as another, you know, another string to their bow, right? They're another part of their decision-making process. It's not that I think coaches have all of a sudden changed their mind and decided, hey, I need AI to make all my decisions for me.

22:03Ryan Beal:But certainly there are organizations that want that information to be provided to a coach to help in his clinical

22:10Gabby:clinical decision-making process but of course it will still only be a guide ryan won't it you know if i look i don't know tottenham manchester united at the weekend that you could have both coaches could have simulated that over and over again look to each other's tactics but ai can't tell you that in the first 90 seconds lamons is going to let a back pass under his foot which leads to a corner which actually united defended but if if they hadn't and they'd scored i mean you you it's a guide is what I'm saying. It can't legislate for, you know, players having a... Wattara diving? Or a dive, yeah, or Ibrahim Kanate bizarrely trying to flick the ball over his head rather than just head it away for Harlow's goal.

22:54Gabby:It can't legislate for that.

22:56Ryan Beal:It can to an extent, right, in terms of thinking about some of the randomness. But there's patterns of decision making and behaviour that we can track through data, and then we can start to understand the likelihood of, you know, of certain players within certain environments to be able to be more prone to, let's say, an error. So I think things like that we can quantify. And the idea with analytics or AI is not to be 100 % right. It's to be less wrong, right? It's trying to get less and less wrong about our decision.

23:29Stephen Smith:And presumably there are certain areas of the pit, well, set plays you can be a lot more precise on. Is it just a coincidence that it feels this season obviously is the rise of set pieces and goals from set pieces is coinciding with the rise of the use of AI?

23:41Mark:I would say set pieces have more been driven by the sort of experts who've come into the game I think Aston Villa aren't you supposed to help them? Yeah they're processing data I think actually from a data perspective set pieces are really hard to collect the data because if you think about how bundled up a set piece is it's hard to then break that apart and the granularity you need when you're extracting that from video data and those type of things to actually look at body poses and all the sort of tactics that go there but I think it's been the rise of the expert in that system which is really interesting from ai because i think the way ai will now go is moving towards more expert systems and what we call agentic ai which is where we're using specific ai systems for specific roles where they can sort of go out and act on certain tasks and in sort of everyday life we'll see this as well where if you want to book a flight you'd be able to use an ai agent to track the flight prices and then be able to just book it and it'll be able to just execute on that so you build these systems that are really good at one thing so then you move away from chat gpt which is very bad at things i think we're speaking before the pod around yeah chat gpt not knowing who players play for yeah you asked me if i use it for work and i said

24:42Stephen Smith:at the beginning of the season i had an experience where i asked chat gpt yeah where the best the key battles were going to be between spurs and villa real and it suggested players who'd left spurs last season yeah yeah so i didn't trust it and that's to do with the training data worse than

24:55Gabby:that it sent me to a pizza express on sunday night in london that hadn't been there for two years

24:59Stephen Smith:i mean you know that's way worse that's way worse yeah definitely glamorous showbiz life

25:05Gabby:on the set piece coaches that i mean honestly emmy buendia scored a direct free kick on sunday for villa and the camera panned to austin mcphee the villa set piece he is not responsible for that but i like he is not responsible for the direct free kick going you mentioned around the the wider of stuff that it can be useful so you talked about fan engagement within the game so will it eventually improve fan fan experiences then yeah definitely i think it's going to bleed into every

25:38Mark:aspect of life i think so it's going to come into that fan experience side and giving us that more of an ultra personalized experience of how we we watch the game but it sort of covers any problem in sport i think when we've been looking at this as a business we've sort of mapped out all of the problems that go through sports and we've got a great board of people who've sort of led some of the biggest clubs in the country and we're looking at problems like athlete abuse and the abuse that they're receiving on social media and we've built tools that can go online and find that abuse and ultimately act on the athlete's behalf to hide it remove it and sort of do all of those things to protect the athlete's mental health so there's so many different applications of it that's being used right now yeah yeah that's being used we used it with the lions over the summer and we've used it with a few different clubs as well to be able to yeah just help so that they can continue to

26:20Stephen Smith:have that engagement without fear of the kind of abuse that they currently have yeah i think

26:25Mark:social media is so important where athletes are sort of almost forced to be on there especially in sports like college football in the u.s where the only way they can get paid is through nil deals and being online and building a profile yeah then you're opening yourself up to this abuse that they can go through and we've tried will that translate to the wider market yeah definitely this can be used in celebrities politicians as you can imagine anyone who's receiving abuse can then use this to sort of remove that and just basically it's a sophisticated kind of a block yeah exactly yes looking at sort of that context and being able to remove that in real time so you just never see it and if there's a threat then you're still able to react because

26:56Stephen Smith:the ai is doing that for you can we focus on women because i'm reading a book at the moment called ultra women which is about the ultra athletes and uh you know something i think we all know there's just been a lack of research into female sports health basically and why injuries occur and there's all kinds of ideas of why they're only increasing acl for example in women in football but that seems to be kind of shifting to other ideas. With this confusion around women's medical needs in sport, how can your AI get the right information when there aren't enough studies in the first place?

27:29Ryan Beal:Yeah, I'm glad you brought it up because I think it presents one of the biggest risks that we deal with with AI as well, that if the contextual information that we're dealing with is not strong enough or if the models are being trained and on potentially the wrong populations, we undermine the fairness of sport right we undermine like the the integrity of the games that we're playing and so if we were to build a model of let's use acl since you brought it up as an example so if we were looking at um every acl injury that happens within our data set over the last x number of years and we look at what are the biomechanical factors that contribute to that what are the physiological parameters that contribute to it the physical parameters the recovery parameters, the training load parameters, et cetera, we could come out with a model that helps us to have a really significant impact on the reduction of ACLs and some recommendations based on that.

28:23Ryan Beal:And then we go and deploy that. And we forget that actually that model was trained on all of our data, not just on females. And we try to deploy that for females. It doesn't take into account that their skeletal frame is completely different. It doesn't take into account that the hormonal release for female athletes is completely different, doesn't take into account where in menstruation, what phase of menstruation are they in right now. And for all of those reasons, we provide them with inaccurate insights, and we may provide them with insights that actually are more dangerous for their current state.

28:59Ryan Beal:And I think that that's a really, really important debate to be had on how we look at how we think about building these models for specific populations and specific groups that's not just females but also for youth athletes or for athletes of different ethnicities like there's lots of different considerations that that need to be given to how we how we deal with data how we deal with building models and how much um we and how well we communicate those limitations to practitioners that are going to leverage them as well because the last thing that we want to do in any of this is to undermine the decision making The idea is to improve it.

29:35Ryan Beal:But I do think with female athletes in particular, it probably has the opportunity to accelerate research. Because you're right, Gabby, there's a huge lack of research right now. But we're getting more instrumented in female sports as well. The amount of data being collected on the physicality of them, on their training load, on their recovery, even on menstruation itself. We're getting a lot more teams that are actually focused on collecting information about the menstrual cycle. but we also know that there's a lack of understanding of what to do with it so if ai can start to pull those pieces together and help us to develop insights on research much faster than we could in a traditional research setting maybe that allows us to learn much quicker within female sports and have a better application quicker just a uh a final one for me is it and

30:20Gabby:it's a bit crass is it expensive will it mean that the haves will increase the gaps over the

30:28Ryan Beal:have not i think the opposite so right right now and i think ryan mentioned this already who did he mention liverpool man city right like clubs like that right okay with brighton etc but they're spending money in in certain areas right but it's all of the big clubs that got mentioned why because they have more money and they have hired what they hired physicists they hired a team of like of of graduates from harvard ai is actually democratizing the entire industry i think this gives every player a chance. I think this levels the playing field. And it means that all this computational power that's been built by these huge companies that we can sit on top of, and we can access and leverage that tooling means that we can make it so much more affordable than hiring an army of data scientists and analysts that come in.

31:12Ryan Beal:And I think that you allow the underdog a chance. I think it actually is probably sport is one of the most beautiful places for it, because that's the beauty of sport, right? Is that everybody should have a fair crack.

31:24Mark:yeah definitely i think we'll start to see it bleed down the game and sort of come into the lower leagues as well i think even this week reading appointed the first i think he's the first head of ai at any professional club so you're starting to see that go into league one and there's plenty of clubs in league one and league two are starting to do things smarter than before i think talent is expensive because you're competing with meta and google to get the best ai talent but actually as you see more of that come into the game and you've got those leading examples of how it's being used at the top, you'll start to be sort of, yeah, being able to use these models.

31:57Mark:They're always getting cheaper. I think in the last year, it's come down sort of exponentially with DeepSeek and Google having their own models and yeah, OpenAI, obviously. So you've got all of these companies competing to train models, which is great for us as consumers and sport because you're starting to see AI prices come down.

32:12Stephen Smith:And then it'll soon, Mark, it'll be your Sunday League team. They'll have their own AI.

32:18Gabby:Yeah, why not? Why not? Maybe the team in Withershaw with all the ex-pros, maybe they'll be the first Vets team. I'm still not sure it's going to get me on social media, though, although that wasn't very beneficial.

32:31Stephen Smith:You don't hear the Hackney Marshes argument anymore, do you? That was the big argument against VAR at the beginning, wasn't it? But can they use it on Hackney Marshes on a Sunday morning? Can you use AI? Well, you can to a certain extent, I guess.

32:42Laura Harvey:The Sports Agents.

32:47Laura Harvey:The Sports Agents. with Gabby Logan and Mark Chapman.

32:53Stephen Smith:So the thing that we didn't really get into, we touched on it a couple of times, Chappers, is that in all of this are people breathing humans with emotions and daily lives that they're kind of navigating. And kids, I've just read the... Give me a clue. Jimmy Anderson's book, Finding the Edge, and there's a brilliant chapter where he's playing at Lourdes against India. It must be around 2011, I'm going to say. He's in a hotel near Lourdes with his wife and kids. His kids at the time, his two daughters were two and one. They all had fevers. It was a hot, sticky London summer's night. His wife wasn't well.

33:34Stephen Smith:He sent her into the next room. They must have had a suite of rooms. Then he spent the night with sick kids. He had to get up in the morning. He didn't even get up. He didn't sleep. Go in and perform and try and take wickets for England at Lourdes. And he did. He took a really great wicket at the end of the day that, you know, in the context of the match was pivotal. AI can't legislate for that, can it, for those days? He said when it was chalked up on the board at Laws, he wanted to, in brackets, have mitigating circumstances.

34:05Gabby:No, it can't. And I suppose that's kind of what we were trying to say in certain areas. I suppose what it is doing is trying to increase the probability of the result happening that you want to. But it will never. That's why I use the Lammons example. It's never, ever going to. And nothing came of that, obviously. But surely for all the tactical influence and so on and so forth, it can't legislate, in my opinion, for the individual errors. or let's go the other way, the individual super performance of maybe someone in the opposition. You said the beginning of that game.

34:49Stephen Smith:What about the end of that game as well? When Spurs should have won that game, but the emotion of that stadium that day and the atmosphere, the highly charged atmosphere, all of that feeds into decision making.

35:00Gabby:And if you look at recruitment, you will still need that human element where you can look at all the data, all the probabilities, all the different lineups, all the different combinations, all the different partnerships, but you still need to go and meet someone and look into their eyes. And let's say you are signing someone from the Bundesliga. Are you going to settle in this country? What is it about you as an individual that means you can move from country X to country Y and be able to deal with that and take that on and be in this part of the country or that part of the country and all the different regional differences there are, you know, would you as a player find it easy to settle in Manchester or are you only going to want to go to London?

35:46Gabby:All of those kind of things. I am guessing that AI can't take all of that.

35:51Stephen Smith:No, and it's important to say, obviously, that, you know, without wanting to sound kind of, we're being a little bit cynical here, that obviously it's going to bring together lots of information and clubs that have already used very sophisticated models like Brighton and Brentford are reaping the rewards of their scouting system that has been along these lines for a while. But the idea that it can filter down to the very lower tiers of football in the same way and therefore is democratic, I'm not so sure that those clubs are going to be able to afford Harvard professors, for example, to disseminate that information.

36:22But I suppose what it could do if you are, God, Shrewsbury.

36:28Gabby:God. Well, no, no, no, no. I was just trying to, I didn't mean, God, if you're Shrewsbury. I meant I was just trying to pick somebody around. So if you're Shrewsbury, so let's think of the markets you might be looking in. And this is National League, National League North and National League South, which they all may be looking in. But then go two or three more steps down the pyramid. And what could you find there? And what can AI help you with there? So it goes back, I suppose, to increasing the sample size and then being able to filter that larger sample size to make it manageable to you. So I can see that argument that it can work.

37:02Gabby:It just needs to be employed in different areas.

37:04Stephen Smith:I think it's a train that is kind of going at speed at the moment, isn't it? And it's not stopping. We're not going to pull back AI. But like everything where AI is interfering with life, it's how it's managed, how it's used. And it's going to be fascinating to see.

37:18Gabby:Well, make sure you subscribe or follow. And please leave a review. You can catch up with all of our past episodes on YouTube, Global Player or wherever you get your pods.

37:27Stephen Smith:And remember, you can always get in touch with us, the sports agents at global.com or messages on social media or via the link in the show notes. thank you for listening that was actually you that said that wasn't it not AR

37:40Gabby:you'll never know this has been a global player original production