Loading summary
Transcript120 lines
- [00:01]
eBay Advertiser
With 132 million users and an unrivaled selection of sports trading cards and memorabilia, eBay is the home of collectibles. Premium trading card singles, live card breaks, unique game, used memorabilia. All the epic finds you need are right here on ebay. Collect with confidence by downloading the ebay app on the App Store or by going straight to eBay.com TheAthletic that's eBay.com TheAthletiC start building your collection today.
- [00:32]
Intuit TurboTax Advertiser
This is a message from sponsor Intuit TurboTax Taxes was getting frustrated by your forms. Now Taxes is uploading your forms with a Snap and a TurboTax expert will do your taxes for you. One who's backed by the latest tech which cross checks millions of data points for absolute accuracy. All of which makes it easy for you to get the most money back guaranteed. Get an Expert now on TurboTax.com only available with TurboTax Live full service. See guarantee details@TurbotaX.com guarantees when you think.
- [01:03]
eBay Advertiser
About red hot businesses like Allbirds or Skims, they have a lot in common. A great product, a cool brand and brilliant marketing. And a business supporting them along the way to make selling simple. That's where Shopify comes in. Nobody does selling better than Shopify, home of the number one checkout on the planet and the not so secret secret with shop pay that boosts conversions up to 50%, meaning way less carts going abandone and way more sales going through. So if you're into growing your business, your commerce platform better be ready to sell whenever your customers are scrolling or strolling on the web, in your store, in their feed, and everywhere in between. Businesses that sell more sell on Shopify, upgrade your business and get the same checkout that those businesses I mentioned at the top use. Sign up for your $1 per month trial period at shopify.com theathletic all lowercase go to shopify.com theathletIC to upgrade your selling shopify.com theathletiC TheAthletiC FC Podcast Network.
- [02:11]
Ayoa Kimulera
Hi everyone, Ayoa Kimulera here and I've got to tell you, we've got a little something special for you today. Ever thought about how elite football clubs actually use data to analyze and recruit players? Well, the Athletic FC Tactics Podcast has the answers you need. Coming up is an interview with Paul Power, the director of AI at data company Skill Corner. They discuss how metrics are used in player analysis, how other sports have influenced football, and the future of AI in analysis. Don't miss an episode of the Athletic FC Tactics Podcast. Listen for free wherever you get your podcast from.
- [02:59]
Michael Bailey
Hello and welcome to the Athletic FC Tactics Podcast. I'm Michael Bailey. This week we have a special guest with us, Paul Power from sports data company Skill Corner joins us to chat about how clubs really use data in recruitment and analysis, plus what the future could hold in terms of football's relationship with AI. Paul, thank you so much for joining us.
- [03:20]
Paul Power
No problem. Thanks a lot for having me on.
- [03:22]
Michael Bailey
It's great stuff. Joining me and Paul, we also have Michael Cox.
- [03:26]
Michael Cox
Hello, Michael.
- [03:27]
Michael Bailey
And Mark Kerry.
- [03:28]
Mark Kerry
Hello, Michael.
- [03:29]
Michael Bailey
So it was September when we recorded a podcast, diving into the world of data, the history of its use across sport and football in particular. But there was one question where we only scratched the surface, that being how much data has truly impacted the action on the pitch. So who better to help guide us through all of that now? But, Paul, let's start at the beginning. Paul, we've mentioned Skill Corner a few times on the POD this season, so we know a bit about you. But it would be great to hear, in your words, exactly what it is that Skill Corner does and your role in it.
- [04:02]
Paul Power
Yeah, sure. So Skill Corner has been around for a number of years now. Basically, the main aim is to allow clubs to reduce their risk and buying players. So that's our aim. So we do this through providing physical data. So one of the biggest issues that clubs have is that in their league, they might be able to get the distances somebody's ran, but if you want to buy somebody from Germany, from Holland, that doesn't exist. So what we've been able to do is to solve that through AI, blah, blah, blah, to be able to detect the movement of players, get the physical data, and then we can provide that to teams so they can actually compare directly players from their leagues to other leagues. In addition to that, we're also trying to provide the most detailed context that you can get about a player. As we all know, the distance somebody runs doesn't tell you everything. How they do it is massively important. And then that also influences, you know, how you might bring them into your team and whether they're suitable or not. So that's a rough nutshell of what we do.
- [05:04]
Michael Bailey
I mean, the bit where you said, blah, blah, blah, maybe jealous is probably the bit I'll come back to in a bit. But I'm also really interested in your, in your background, how you sort of got to this point.
- [05:13]
Paul Power
Yeah, sure. So. So my current role now, I'm director of AI at said Skill Corner. So my role is Specifically around, you know, developing this contextual information about a player. So the type of movements they make. Do they press, how do they make decisions? But before that, my background was actually more coaching and sports development. So I worked for Sunderland Council as a senior sports development officer, building swimming pools, stuff like this. And then I used to do a bit of coaching on the side. The big, you know, financial hoo ha happened. People had to be laid off, so gave me a chance to go and retrain. But at the time, I was actually following Michael's zone and marking blog, so that inspired me to start writing about tactics. Kind of got into that a bit more. And then you actually suggested Moneyball as a good book to read.
- [06:01]
Michael Cox
Yeah.
- [06:01]
Paul Power
Remember this? So it's like, give it a go, read it. I was like, wow, you know, this totally blew my mind. And that probably one moment has led to me getting to this. So I went to Sunderland Uni, retrained, learned about dynamic systems and chaos theory and how this applies to sports. And then from that, I joined a company called Prozone, who I think a lot of people in the football industry know they were the first company to really bring tracking data and physical stats to the game. Really famous with Sam Allardyce and Steve McLaren. And then from there, I've kind of gone on and gone on. Probably one of the first data scientists to work with tracking data to try to build these insights rather than just say how far people have run. And then that's led me to this conversation here.
- [06:41]
Michael Bailey
Look at what you've created, Michael.
- [06:42]
Paul Power
So it's all your fault.
- [06:44]
Michael Cox
Yeah, I mean, I think the Moneyball thing is really interesting because I think the podcast we did before was. Was timed to. To go along with the article I wrote at the time about Dayton football. And one of the points I made in that is I think Moneyball, the book and the story, is almost more famous than the team involved. I don't think, you know, unless you're a big baseball fan. I don't know anything about the Oakland days, but everyone knows about Moneyball. And I think within data, it's almost been like the race to tell the story has been as keenly contested as the race to actually crack the code, if that makes sense. And it's quite interesting that last couple of years, all these books about analytics, a lot of them are written by the people involved, and a lot of the other books would focus on those people as part of the story. And then we've had their story. So that's what I've been quite intrigued by. With the whole data thing actually telling the story and working out to a certain extent trying to separate people who are incentivized to say that data's had a huge impact and of course, looking at how big an impact it's actually had.
- [07:41]
Paul Power
Oh, it's had a huge impact as someone who's incentivized to say that, but I do think it's a really great, great points, you know, you're making. I was listening to the podcasts and it's quite a difficult thing to really pull out even, you know, so I've worked in the industry for 12 years now, nearly at the beginning of tracking data and even now it's really hard to pin down and say, you know, like in basketball, the three point thing is such an obvious aspect to award to data, but it's less obvious in football. But I do think there are some areas.
- [08:14]
Michael Bailey
Can you think of exact examples of where maybe data has influenced what actually happens on the pitch and some of the decisions that are made there by footballers or maybe in the coaching and development of tactics in terms of what's happening in games as well?
- [08:26]
Paul Power
Yeah, so I think obviously you mentioned reap in the, in the podcast and you know, you go back to that and it's a very obvious one, but it did, you know, kind of ruin football, you could say for a while in multiple countries. You know, obviously England, even if you look back to the DNA of England, which is obviously a, a popular thing that we speak about now, but back then the coaching manuals weren't about going long. So like I did my coaching badges, so I've got my B license as well. And you could see that when you were first starting on your level ones, level twos like 25 years ago, showing a bit about my age and you were coached about direct football, you know, it wasn't about this beautiful ticky tacker then obviously you see it in Scandinavia, you know, Sweden, if you read a lot of that, they've got a lot of great guys who look at dynamic systems and about the social impact of a country on how, on a team's identity and they still, you know, talk about Hudson's impact of being very direct, you know, playing there and that's, you know, still part of them and it's come back. So I do think that's had a lasting impact on football. But what I would say is that I think data hasn't directly influenced how football is played in terms of, hasn't just suddenly come up and say should play this way, but what it's done is confirmed, you know, perceptions. So, you know, I think there are two big reference points. One is, you know, Pep Guardiola obviously, you know, kind of restarted football, you could say, but the data backed up what he was doing and you can see it now, you know, even down to like league one, league two in England. This is how managers are recruited, you know, so while it's not necessarily they've said you should play this way, it showed that that's a success. So when managers are being evaluated, they have to meet certain key criteria. So I do think it's had a big impact on that. And the other aspect is the counter pressing side. You know, it's, you know, being validated through, you know, Pep talking about you had five seconds to be able to counter press to win the ball back. And I did the research on this many moons ago and the actual answer was 5.3. So it was a pretty good guess from him. But we were able to demonstrate that if you counter pressed, you were able to prevent a dangerous counter attack if you were able to apply pressure within that time kind of over that, you know, and obviously have plus or minus one or two seconds. But the probability of conceding was really high. So I think that there are these things that have been seen in the tactics that the data is then backed up. But it's not that the data is then, you know, created the tactic first.
- [10:56]
Michael Cox
Yeah, I mean, I think that's, that's the most interesting thing. And I actually think one of the good examples of that is an article Mark did which was about the change to the goal kick law in terms of you could pass it inside your own box. And Mark found, you know, almost every Premier League team is trying to do that and take advantage of that. And Mark found that actually teams were progressing the ball further up the pitch when they played the ball short rather than going long, which I think was a good example that it probably didn't come from the data. The data shows it was the right approach.
- [11:22]
Mark Kerry
Yeah, it just reinforces, as you guys say, it reinforces the approach that you're taking tactically. And I think the cause and effect is really interesting there as well. And obviously we know that data has kind of bled through so many different departments within a club as well. And I suppose even just down to things like simple match reports after a game as well. Just saying, did we, you know, even if we lost a game, are we still trying to do the right process? Are we still on the right lines? Like, it, it just gives a window into then what you're doing on the pitch tactically and reinforcing it. If a manager is struggling, they'll maybe sometimes use it in a press conference as well and throw maybe something as simple as an expected goals metric out there to say that the process is looking good and here's why. So I agree it's maybe not always the cause, but it's a really good tool to help.
- [12:05]
Michael Cox
So what are managers actually looking at? I mean, I don't think no one really disputes the idea that you can measure football quite well through numbers. And in terms of recruitment, just why wouldn't you use numbers? You've got database from thousands of players. But if a man just looking at his own team's performance, whether he wants to confirm suspicions, whether he wants to find things out, what metrics is he looking at beyond expected goals in terms of how's my team played today?
- [12:31]
Paul Power
Yeah, so every club will have certain KPIs that they want their manager to meet. So that's at the ownership level, you know, director, but then the manager as well will usually ask for certain KPIs for every match analyst that they want to track. So either that's through, you know, stats that we provide or other people but stuff that they'll manually collect as well. While watching the game, what would be.
- [12:53]
Michael Bailey
Good examples of those KPIs? Could you just think of one of like.
- [12:55]
Paul Power
Yeah, like the simplest one is final third entries, which is such a simple metric. But when I was consulting to Everton when Martinez was there, and that's what they really benchmarked themselves on because obviously you've gone from a David Moyes style team to want to become a more progressive. And what we're able to show is you were very good at getting into the final third and the routes they went. But then it was about time to shot. Once you get into the final third, how long did it take you to break them down? And that's initially under Martinez where they struggled so they would dominate the ball. But you compared them to the other big teams, so that's what they would always want to see. What we like compared to the top four, how far off were we? And they liked this big outlier. So there was good in possession in terms of getting it there, but then to convert it is where he was struggling.
- [13:43]
Michael Bailey
This is fascinating because you can't as a player, you're not going to be on the pitch going, oh, hang on, no, I've gone on too long, what am I going to do now? You know, the head coach has got to coach a way of getting those data to improve which is like such a challenge in itself, which is what.
- [13:57]
Paul Power
He, you know, Martinez did really well, you know, in that first season when they, they came fifth, you know, in the last minute, you know, they were really able to remold Lukaku, how he played. So that's where he went from being central to out wide. So again the data was showing that's what he's most dangerous. So when you gave him the ball on the right, he was able to cut in. That's what he was going to get. His, his goals. If he was in the middle, you know, back to goal, he was surrounded. They couldn't get it into him, they didn't have the runners so that meant they couldn't break them down. So we were able to use these relatively simple metrics. They were then able to evaluate what was going on and as they saw it shift, they saw the league position go up as well.
- [14:36]
Mark Kerry
I suppose that's what's interesting. It's something I bang on quite a lot about with the difference between data and then impact basically. So it's going from a bit of data there in the final third entries into then maybe into information, into insight. Okay, well why is this happening? And then impact is then changing the like tactically changing initially a piece of data into then what happens happens on the pitch. So you can see the sort of the timeline and the trend there from actually using numbers to then. Yeah, on pitch.
- [15:02]
Michael Cox
I mean I think one of the most interesting things Deja can do in terms of individual players is very closely related to that about Lukaku which is strip away some of the biases we might have. And I think Lukaku has always been slightly typecast because he's so big and so strong and therefore I think people look at him and think he's a proper penalty box number nine. But I think he's a lot closer to a Terrier Henry or even a Mohamed Salah in terms of being left footed on from cutting inside from his, from his right foot. He's never been particularly good with his back to goal. I don't think his link plays great. He can hold the ball up okay. But yeah, I mean it's interesting that's the one you mentioned because he's always been a player where I've been, you know, whenever he's done well for Belgium in the World cup against Brazil, there was that quarter final, I think when he was excellent from the right for Everton when they think they're playing Naismith, was it Naismith up front?
- [15:54]
Paul Power
Stephen Naismith, really intelligent player, really good player.
- [15:58]
Michael Cox
And even for Inter playing in a front two. So he's always just on that side of the pitch. So yeah, I think sometimes data can strip away the visual aspects of players that we might compromise our judgment.
- [16:11]
Paul Power
Which interestingly goes back to the money ball because that for me is a large part of it is, you know, like Chad Bradford, isn't it? You know, the classic one where he throws funny, you know, and we all do it, we look at somebody. So Mark, I looked at you and my first thought was bet you were a good center back, you know. But we do, we have these innate biases and you know, that is the beauty of data. And we have requests. So there's something called body pose which you may or may not be aware of. Those who aren't aware is where you can start to like track the skeleton of players. So we're able to do that now, which is really, really cool. A. It looks fancy but when we've spoken to some clubs, they actually want us to generate full matches of just this skeleton data. You don't know who the player is so that the scouts can watch it to strip away all bias. And there's been some research on this by I think Sam Gregory, who's Now at the U.S. men's National Team and he did. They were able to demonstrate that, you know, when you watched it and they had men's and women's games in as well. And when you just watch the body pose, the scouts couldn't tell as soon as he watched the video straight away, the biases crept back in. So I think that's something in the future that could start to influence scouting, which I think is pretty interesting.
- [17:25]
Michael Bailey
That's so interesting. We will come on to that later. I am quite interested in how so much data is collected and how that has maybe changed skill corner use television cameras, don't they? So that is that just one camera? Is that all the angles? And how do you know where a player is if they're not on camera?
- [17:43]
Paul Power
Yeah, so that's the magic. So the way it works for those who don't know, we take what's called in the industry a single camera feed. So we take the broadcast television footage. So we have no cameras in stadium, so we can recognize the players through AI or computer vision. So we know who they are, what the number is, etc. But as you pointed out there, there's only like 15 players on the screen at once. So what we then have to do is to, is to simulate or predict where we think the other players are on the pitch given the dynamics of how everybody's moving. So it's ridiculously cool what the guys have solved, but it allows us to then have this full picture despite only having this narrow camera view.
- [18:28]
Michael Bailey
Last week we did speak on the POD about the difficulty in assessing defenders using data. We went into that in quite details. How have you found that? Analyzing that not skill corner necessarily, but just in generally, Paul, in terms of that it's easy to register and look at the pretty obvious analytics and metrics. But in terms of defensive side of the game, how tricky has that been?
- [18:53]
Paul Power
Yeah, it's probably has been the most difficult area to evaluate. So I think you guys said it before, it's like the absence of action. So I did some research when I used to work for Stats Perform. It was looking to try to identify the influence of a defender without touching the ball. So can you prevent what's happened before it's happened? Because that's really what a good defender is. You know, that's, you know, you always go back to the Van Dijk, you know, if this area, you know, you've got to talk about Maldini, Beresi, you know, all these guys. But that's now where we're starting to get to. So we've, I wouldn't say we've solved it, but we're pretty close to being able to identify this now. So just understanding, you know, a defender has stood in a position, the ball doesn't get played to their attacker. But we can start to understand that's down to the influence of that individual. But then also so you can start to reward that to each individual player rather than just the tackle that somebody's made. So I think that's where we're going to in the future. So it's what's not happened. Evaluating what's not happened rather than evaluating what has happened, which I think is.
- [19:56]
Mark Kerry
Interesting as well, because you can look at what has not happened when a player is in possession as well, because you can look at all of the options that they have and you think which players may be always taking that easy 5 yard pass versus actually there's a teammate who's showing for the ball and they've elected to not go for that more tricky pass because maybe they want to play it safe or whatever. You guys can do that. And that's growing as well, isn't it? In terms of decision making in possession with the options that a player has in Their viewpoint, I think that that's.
- [20:21]
Paul Power
The key differentiator now is that yes, you knew a pass was made, you might know a dangerous pass is made, but if somebody makes a backward pass, you kind of get punished for that normally. So there's something called like a EPV or expected point value for those who don't know. It's basically for every action I make, do I increase or decrease the chance my team scoring. So you pass the ball backwards, it's always negative because it's backwards. But that's not the right way to evaluate because if doesn't have an option to go forward, why would you penalize somebody for playing it backwards and keeping the ball? But again, you should reward the defender because they prevented everything from doing that. And that's now where we're able to get to is, is to see that complete picture.
- [20:59]
Michael Bailey
Paul, you mentioned earlier, Reep, that's in terms of Charles Reap, isn't it? Michael?
- [21:04]
Michael Cox
Yeah, in a way he was probably the first analytics person in English football. We're going back to the 70s now and he was doing a lot of, I mean in a way quite advanced things in terms of not too dissimilar from expected goals and expected threat, but kind of read the data in the wrong way and basically prescribed long ball football and getting the ball into the box within five passes. And yeah, Jonathan Wilson's Inverting the Pyramid is a very good source if you want to know more about how that slightly derailed English football for a couple of decades.
- [21:44]
eBay Advertiser
Ever feel like everything's getting more expensive? Groceries, rent, even breathing? And let's not forget car insurance. It just keeps creeping up. But here's the thing. Insurify makes one part of life easier. It's an insurance comparison site that finds real time quotes from dozens of top insurers. Insurance companies compete for your business, driving down prices. It's like having a friend who knows all the best deals. But without the smug grin. You could save as much as $1,025 a year. And you might even cut your bill in half. The best part? It's all online. No phone calls, no spam, no nonsense. You could be covered in as little as 10 minutes. And with an excellent star rating on Trustpilot, saving feels less impossible. So why not give it a shot? Visit insurify.com and see how much you can save. That's I n s U-R-I f y.com average potential savings based on initial quotes received by 183,234 customers seeking insurance through Insurify. Actual savings may vary depending on state of residence, individual circumstances, coverage selections and insurance provider savings. Results are not guaranteed.
- [22:49]
Home Depot Advertiser
During President's Day Savings at the Home Depot, get up to 35% off select appliances plus save up to an extra $450 on select appliances like Whirlpool Shop, the side by side fingerprint resistant stainless steel refrigerator with an indoor can caddy or the fingerprint resistant stainless steel top control dishwasher with three 5 inch shelves. Make the most of President's Day savings and our lowest prices guaranteed at the Home Depot. How doers get more done. See homedepot.com Pricematch for details. See Store Online for details. Minimum purchase required.
- [23:20]
McAfee Advertiser
Your data is like gold to hackers. They're selling your passwords, bank details and private messages. McAfee helps stop them. Secure VPN keeps your online activity private. AI powered tech scam detector spots phishing attempts instantly and with award winning antivirus you get top tier hacker protection. Plus you'll get up to $2 million in identity theft coverage, all for just $39.99 for your first year. Visit McAfee.com, cancel anytime terms apply.
- [23:56]
Michael Bailey
We've had a few questions in from our wonderful listeners, Paul, so let's fire a few of those at you. Mark Thompson let's start with his we occasionally see distance run or sprints used in the media as indications of effort. If you wanted to motivate a team to give more effort using a physical running metric as a target, what stats would you advise?
- [24:18]
Paul Power
Yeah, it's a, it's a really good question because measuring physical data and then telling somebody how far they've run is the simplest thing you can do. And you always hear that when a new manager comes in, it's like these guys aren't fit enough. Well, no, I'm sure they possibly are, but I think the reality is now the industry is moving away just from distances and it's about quality of movement. So now it's really about accelerations and decelerations and explosiveness. And that's really what they're evaluating their players on now. So you might go into a transition situation. So what's your reaction time? Basically to be explosive to try and win the ball back. And these are now the critical battles. You know, just covering, you know, 10k in distance or over a thousand meter in high speeds, which is considered like a gold standard benchmark, is just standard now. But I think for me now, if I was working in a team, particularly with some of the brains that are in clubs. I'd be more interested in identifying what distance does a player prevent from happening. And it kind of goes back to the Pep Guardiol and the whole, you know, counter pressing. So if you're all out, attack the opposition, have a counter attack. What does your intense effort prevent, I. E. Being able to get to you, stop you, force you backwards. Which then means five players don't have to do a 30 meter recovery run. And I think that type of thing is cool because you're rewarding the individual for sacrificing himself for the team. And I'm not sure if teams are doing this but you know, I think that's where we will get to.
- [25:54]
Michael Bailey
I love that because I was in full on cynical mode of just envisaging lots of players just running around really intensely to look like they were working really hard. But that's just me being cynical, Mark.
- [26:04]
Mark Kerry
No, no. Historically players have done that to get their numbers up from the, you know, the GPS vest that they've worn. They've gone to sprint to go and get the ball if it's gone out for a throw in and stuff. They've been like the age old examples, haven't they? But I think the contextual factors there are crucial to the extent that actually you could as a team run less and the output that you're saying would actually be rewarded more.
- [26:23]
Paul Power
Yeah, because there is actually no direct correlation between running further and winning. No, you know, we know this but it's about how you do it. And we've generally found if you look at your matchups, the team, so if you have attackers first, defenders, if the attackers are performing more intense actions, then they're probably going to win the battle as a result, win the game. But you say there about guys, you know, trying to inflate the numbers. When I was at Sunderland, that is when GPS vests first came about.
- [26:50]
Michael Bailey
This is the football club, not football club, sorry.
- [26:52]
Paul Power
Yeah, yeah, I used to coach with the community but I was analyst with the first team in the academy as well. And you used to have guys like literally jogging on the spot or jumping up and down and like the guys, sports science is like, you know, what you doing? He says, well, you know, getting me numbers up. He's like, yeah, you stood still, you know, so. But in their head they were like, yeah, I'm getting me numbers up here. So yeah, that definitely has happened with people trying to game us.
- [27:16]
Michael Bailey
Love it. Carlos has been in touch as well. Throw ins are by far the least regarded set piece. He says, I don't know if that's an official statistic have you seen any teams using throw ins in an interesting way or is that an area that's been picked apart a bit?
- [27:31]
Paul Power
It does not through our data. You know I won't lie but you do see there is some really interesting research now about the value of throw inside but not necessarily creating but from not losing the ball and you know that's the way I always looked at it as a coach you know you would you know throw the ball but your probability to keep possession massively reduces So I think there has been research into that but I won't lie there's nothing from from our own data points.
- [27:58]
Mark Kerry
That focuses I mean Thomas Granamark is the the one who's held up as the the gold standard coach for throw ins and to exactly to Paul's point it I think he held the record or holds the record for the longest throw of a football or something along those lines and it was everyone who employed him Others thought that it was because they were maybe trying to do more long throws but to Paul's point it was actually trying to maximise the use of throws when it was short throws keeping possession making sure that the throw was accurate enough to actually go straight to the players feet rather than it go into their midriff and they've got to then control it and then lose possession as a consequence so Thomas Grondermark is the one to go to there but I did look into the numbers on this and the across the top four European leagues so the typical top five European leagues minus the French Ligue 1 Brentford and Bochum in the Bundesliga have 19 shots from throws this season which is the highest so look out I mean Brentford didn't make sense we know this but yeah I'd look out for Both of those two Getafe are up there as well with 18 shots from throw ins Friends of the Pod and Nottingham Forest Ahmed Walid did a really interesting piece about Nottingham Forest throw ins of how they not necessarily from a long throw in but the sequence of plays play that they do so they're up there as well with 17.
- [29:11]
Michael Bailey
I think every piece Ahmed Walid writes is interesting I think we always preface it with that I mean having covered and watched a lot of average football I've seen a lot of bad throw ins years and years of them as I'm sure you have too saw some.
- [29:23]
Michael Cox
Last night at Kingstonian We've got another.
- [29:27]
Michael Bailey
One here from Prasanna Patil who has asked this might be quite a long answer Paul, So see how you go with it. If you had to choose one stat to emphasize performance for each position on a football pitch, what would it be?
- [29:39]
Paul Power
Right, so there's 27 positions. So.
- [29:42]
Michael Bailey
Well, not in the brackets in this post, but yeah, go for it.
- [29:46]
Paul Power
Yeah. I think that's one of the hardest things to do, is to distill something down to one number and it's, it's something that's been done in American sports and you can see it creep over now into football. Obviously the simple answer is it depends. I think we spoke about possession value and this expected threat I think is a really genuinely great catch all metric because it does capture process. But you know, for me, if I'm looking at center backs, am I able to prevent a passing option for fullbacks? What's my ability if I combine things together to be able to get to the final third in a counter attack but then be able to recover straight away as well. And speaking with a guy called Michael Beale who was at Liverpool Academy and he was saying these are the stats that they use to evaluate fullbacks. You're up against Gareth Bale in Champions League at the time, you're saying, so you have to be able to get forward, get in the cross, you know you're going to lose possession, you have to get back and match Bale. So it's those type of metrics that you're probably looking at. But I think a bit of a cop out would be this expected threat. And then, yeah, there's a lot for each position and there's a lot for.
- [30:58]
Mark Kerry
So many different teams. I think you started it by saying it depends and I totally agree with that because it depends on the game model, it depends on the remit that the coach or the manager is asking for. So it's an interesting question, but it is one that could be deserving of its own podcast because it really does defend.
- [31:13]
Michael Bailey
Maybe we'll do that one in future.
- [31:14]
Mark Kerry
Yeah, yeah, yeah, that's a good idea actually. Yeah.
- [31:24]
McAfee Advertiser
I can say to my new Samsung Galaxy S25 Ultra, hey, find a keto friendly restaurant nearby and text it to Beth and Steve. And it does without me lifting a finger so I can get in more squats anywhere I can. 1, 2, 3.
- [31:38]
eBay Advertiser
Will that be cash or credit?
- [31:40]
Paul Power
Credit.
- [31:41]
eBay Advertiser
4 Galaxy S25 Ultra, the AI companion.
- [31:45]
Paul Power
That does the heavy lifting. So use you can do.
- [31:47]
eBay Advertiser
You get yours@samsung.com compatible with select apps. Requires Google Gemini account. Results may vary based on input. Check responses for accuracy.
- [31:54]
Paul Power
My dad works in B2B marketing. He came by my school for career day and said he was a big roas man. Then he told everyone how much he loved calculating his return on ad spend. My friends still laugh at me to this day.
- [32:08]
eBay Advertiser
Not everyone gets B2B, but with LinkedIn.
- [32:11]
Paul Power
You'Ll be able to reach people who do. Get $100 credit on your next ad campaign. Go to LinkedIn.com results to claim your credit. That's LinkedIn.com results. Terms and conditions apply. LinkedIn the place to be to be.
- [32:24]
Michael Bailey
Get the Angel Reese Special at McDonald's.
- [32:26]
Mark Kerry
Now let's break it down. My favorite barbecue sauce, American cheese, crispy.
- [32:30]
Paul Power
Bacon, pickles, onions and a sesame seed bun, of course.
- [32:34]
Michael Bailey
And don't forget the fries and the drinks.
- [32:35]
Paul Power
Sound good?
- [32:37]
eBay Advertiser
I participate in restaurants for a limited time.
- [32:46]
Michael Bailey
Please do keep the messages, questions and discussion suggestions coming. We've got a pot of them all and we'll be getting stuck into them in future editions of the podcast, I promise. Simply email tacticspodheathletic.com Paul Power from sports data provider Skill Corner is with us for today's pod. Paul, I like this. Any misconceptions about the data world from the outside that particularly wind you up? So free rant, go for it.
- [33:11]
Paul Power
I think the term boffins has to be up there. So yeah, that one. And nerds is always a great. But you know, I think a nerd for me is somebody who's very passionate about, you know, what they try to do and is usually very good. So, you know, I'll take that as being a nerd.
- [33:32]
Mark Kerry
Yeah. And it's used as an endearing term among our colleagues. The data nerds have helped me with xyz, but it does great on you just a little bit and let that be known.
- [33:42]
Michael Cox
I think boffins is more complementary than nerds. Would you agree?
- [33:45]
Mark Kerry
I say so. Yeah.
- [33:47]
Michael Cox
I don't think boffins is disparaging, whereas I think nerds probably is. Where do you stand on geek similar to nerd? I think.
- [33:55]
Mark Kerry
I think so, yeah.
- [33:55]
Michael Cox
Yeah.
- [33:56]
Michael Bailey
If you could whip up an expected offense metric for all this, it'd be great. It has changed, hasn't it Paul, in terms of how media, but essentially fans as well through the media have kind of accepted and taken data and analytics as part of their sort of football enjoying experience generally. Do you sense there is that warmth or is there a point where it might get, you know, there might not be much more of it and it will reach a kind of a ceiling almost?
- [34:26]
Paul Power
I think it's only going to grow, to be honest the more it gets adopted in the media. So the job that you guys do I think is brilliant the way that you're able to explain things. So I know Mark, I messaged you before when you're explaining like the runs and behinds with the metrics that we do, the more we can simplify it for people to get on board with, the better. And I think that's not just for fans, it's across football in general. Like we have these terrible names about possession value, expected goals. Our original one was called on ball movement points. How sexy is that? Terrible. So I think the more we can simplify things, more fans will get involved. But like even my friends, you know, before it was a Mick take, you know, it's like, oh, Paul, I bet you know, you know this number or whatever. But now like they're actually coming to me and saying, I've seen how many just simple like shots inside the box or somebody will reference XG now. So I think it has become accepted and it's only going to grow.
- [35:18]
Mark Kerry
I wrote about this fairly recently about it was actually on a slot after a Liverpool Brentford game where he used expected goals in his post match press conference. And I used it as a springboard for an article to say that actually for fans who were previously maybe skeptical about the use of it, they now have to keep up with the fact that managers are using it. So if they want to sort of be fully engrossed in their club and the way that they like to approach things, then it's now kind of a keep up or fall behind because your own managers, your own clubs are using it as part of their, their own sort of coverage and their answers. So this is the way that it's going now and those skeptics who hated it on Match of the Day are maybe getting quieter, shall we say.
- [35:56]
Michael Bailey
It's interesting though, because I've been in press conferences where managers referenced XG but in a way where I've had Mark's voice in my head going, you're not supposed to use it like that. And that's from a manager in terms of inside football as well, from people you speak to, I mean, do you sense there's still some reluctance or some questioning from a lot of people? And also we obviously get exposed to quite a lot of analytics. But how different is that to what would be going through a football club and what they would have to be dealing with and processing?
- [36:21]
Paul Power
Yeah, so there is still some resistance, you know, I think there always will be. I think part of the biggest issue is how it's sold. A while back when Strachan, Gordon Strachan was manager, affaired Scottish national team, I went up to. To pitch him, you know, expected goals, possession value, and it was probably one of the best meetings I've ended up having ever. Speaking with coaches. It didn't start off well. He turned out he'd had a medical and basically been told that he had to reduce his, you know, maybe fat foods in his diet. And then I'm stood there, you know, in a shirt and tie, look like a sales guy. You've got Stuart McCall in the room and Mark McGee as well, you know, and they've only just said, you're right, Paul. Yeah. Then Strachan comes in, goes, right, what you going to waste me time on for the next hour? You know, And I'm like, right, okay, this is starting off well. My first video was showing Fish, which he was like, well, what on earth is this about? So I was trying to explain to him about how football can be thought of as a dynamic system and that if you look at shoals of fish, this is used in ecological psychology and stuff like this to measure interactions between agents or fish, in this case.
- [37:33]
Michael Bailey
So if some fish go some way that will affect where the other fish go.
- [37:37]
Paul Power
Exactly.
- [37:37]
Michael Bailey
You're dealing with.
- [37:38]
Paul Power
Yes. But basically just to explain to them, we're trying to analyze football as a dynamic system and if a fullback moves one way, that means the center back is going to move this way. It's exactly the same in dynamic systems. And then we spoke about xg, but showed him the video of Roy Hodgson when he was England manager. I can't remember if it was a friendly against Norway and somebody from a newspaper said, oh, you only had two shots on target, Roy. And if you remember, he went two shots on target, two shots on target. What about when we hit the post? What about us over the top? And then Strachan was like, exactly. I was like, well, that's expected goals, guys. And they were like, like, oh, okay. So we got into that, we spoke about possession value and straight away was like, so you can tell me the process. Even if we make a cross and my striker doesn't get onto it, I can still reward him and I can shout at him. I was like, yeah. And then we got onto something called pitch control, which didn't fully exist then. So this basically tells you, given where I'm stood in the other players, what part of the pitch should I get to first so I have ownership over this area. Something called Voronoi diagrams, Very Sexy. Basically it draws an area around a player and where you expect two players to bump into each other, you draw a wall, basically. So it's like a cell. If you remember at school you looked at an onion cell, you know, in biology and stuff kind of looks like this with a nucleus in the middle, which is the player. And we were showing these Voronoi diagrams of spatial control about Thomas Muller being exploited. If you got the ball here and then we had the coffee cups on the table and the guys just got it. But I think it's because it was explained in their terminology, and that does still feel a bit of a limitation at times, is really communicating these really complex things, because they are complex, but they're not. It's measuring football. And the guys understand that. As long as you explain it in the, in the right way to them.
- [39:22]
Michael Cox
I still think expected goals is just not a very good term.
- [39:25]
Paul Power
It's a terrible term.
- [39:26]
Michael Cox
It's stuck. It's stuck now. You can't go back. But I mean, I saw someone on Twitter last night saying the XG figure from the Liverpool game doesn't make sense because one of the chances fell to Nunez and Nunez always misses. So why is it relevant? But it's obviously just a shots figure adjusted for the position of the shot, essentially, and more, more complexity coming into it. But I don't think expected goals works to someone who isn't from a statistical background.
- [39:52]
Michael Bailey
Personally, you mentioned a little bit there as well, Paul, about other sports. I think I, I've kind of liked the idea of doing a podcast on how football's been influenced by other sports, but I'm. I'm not sure if it has and we've never really kind of got there. But I can ask you the question, like, I mean, you've worked in other sports. Please tell us which ones those were and also maybe how that, how that differs from dealing with football, soccer, and whether. Whether there is sort of influence that can be taken from, from either end.
- [40:18]
Paul Power
Yeah, definitely. So I've been able to work across cricket, American football, baseball, basketball, rugby union, rugby league, obviously football. So I've had a good breadth across these. I think what's interesting, going back to the genesis of analytics and football, a lot of it was inspired by ice hockey. So they kind of created the, their version of xg. And while XG might have been invented, you know, in Ben Houses and Charles Reap, etc in terms of the public realm, it wasn't there. So that really a lot of ice hockey influenced a lot of early analytics stuff called pdo can't remember what it was reference to but basically your shared of shots on target percentage wasn't PDO.
- [40:59]
Michael Cox
The thing that didn't actually stand for anything?
- [41:01]
Paul Power
Yeah, I might be, I might be.
- [41:02]
Michael Cox
Making genuinely, I might be making it up, I think. I'm sure it actually didn't mean anything, just sounded good.
- [41:09]
Paul Power
Yeah, I'm not even sure if it was.
- [41:10]
Michael Cox
Sounds wrong. I apologize. Some will correct me.
- [41:13]
Paul Power
No, it genuinely. Yeah, I think there's some truth. There is some truth in that. But I think what's really interesting is that football is the most mature in terms of data. But apart from maybe baseball, you know, baseball's had it for a long time, but baseball is a very simple sport to model. You know, you basically have two people that you're competing against. You know, you run around some bases. That's not to take away from those guys there because they are phenomenal at what they do. But yeah, so football I think has generally influenced like basketball and has influenced American football initially, but now basketball has then gone back and influenced football a lot. So there's a guy called Luke Bourne who's Adzelas and he's also part owner AC Milan I think now a lot of work he's done. Possession value, they built it in basketball. But now you look at American football, which is probably the slowest to adopt analytics, they've now got tracking, so we work with teams or franchises as well and they're starting to kind of build up their like muscle memory in this. But probably the other big thing that football's influenced is sports science. So you've got football and Aussie Rules football probably leading the way, like my guilty pleasure. Right. So a lot of phenomenal work in that area and then that's now come over to really influence athlete monitoring as it's known, particularly in American football, and that's now linked into ice hockey, etc. So there is this kind of cross pollination across sports to definitely.
- [42:39]
Michael Bailey
Yeah, AFL is definitely one of my guilty pleasures. Paul, you mentioned AI as well in terms of how you're creating some of these models. I mean that does seem to be a bit of an elephant in the room in terms of how AI is going to influence and alter maybe football and analytics going forward. What is the landscape like at the moment and how do you see it changing if you can stare into your crystal ball?
- [43:02]
Paul Power
Yeah, yeah, it's an interesting one mainly because I think you just hear this term kind of evolve. So initially you were a data analyst, then you became a data scientist, then you were working in big data, then you were working in AI, you know, so I think a lot of these things are very similar. I think AI in terms of sport is mainly used to collect the data. So, you know, we use AI for us to do player recognition, etc. To simulate what all the players are in terms of AI, for me, it's about simulating to have, you know, to understand what if you've done something else, it's not there in sports at the moment. People have tried. There's a very famous old article which referenced the Toronto Raptors, where the concept of ghosting starts come about. If you're not familiar with it, if you played Mario Kart as a kid, you know, you race, you know, and then you have this ghost that you want to try and beat. You know, when you do the same circuit, again, really similar principle in basketball, but you were ghosting. If I'm a defender, where would an average defender be instead? You know, would they be closer? Would they be further back? And then at stats, we. We introduced this into football as well. So can you again simulate where defenders should be, given the positions of attackers? But it's never really then taken off into the mainstream to help with the tactical aspect. But I do feel that's starting to get closer now as the data gets bigger and teams are more accepting of using data as well.
- [44:29]
Michael Bailey
That's interesting. I've seen that a lot in Formula one, say for like pole position laps or fastest laps in qualifying. You can see footage of the two cars on the track at the same time. But they're, I guess, both factual because that is what those laps were. Whereas this is more about an average and a kind of a modeled behavior against real behavior as well.
- [44:50]
Paul Power
So, yeah, I mean, I do see. So again, when I was at stats, we created a demo for basketball, particularly like for the, you know, beating shot clock. You know, the guys call a timeout. You know, you normally see the coach draws something with the X and O's, shows the players, this is what we want to do. So we basically took that concept. You know, they draw it out, but you simulate it and then it will then tell you we actually think this is what the defenders will do, maybe recommend something else. So I do feel that's eventually where we could get, you know, particularly if you think of set pieces in football, how you get the information on the pitch is a bit more difficult, but you can kind of see that as the easiest thing to influence.
- [45:28]
Mark Kerry
Yeah, and that's probably a good example of then, yeah, it coming Back from data to more on pitch influence. And I know that Liverpool linked up with Google DeepMind, didn't they? In terms of. I think it was defensive corners, but talking about corners and saying how if they were to augment and simulate a player in a different position, would they have prevented the corner from. Sorry, prevented the shot from happening from said corner. But it's funny that artificial intelligence has become such a buzz term in the world at the moment and people or not everyone, I think really kind of fully understands what it means and I suppose there's different types of it as well, I guess, to your point. And I've spoken to a few companies who are doing more generative AI from a football perspective and using a football version of a chatgpt, essentially to say, can you tell me about this sort of player? And it will use the analysis and the AI to be able to actually reply and in real time you can reply, I think so. There's a company I've spoken to called 12 who you can just engage with them over WhatsApp and say, can you tell me more about why Chris Wood is having such a fantastic season? And it will give you some back and forth, it'll give you some more information if you need it, a report, et cetera. So in terms of AI more broadly, there's kind of branches of it. Is that sort of fair to say as well?
- [46:36]
Paul Power
Yeah, yeah, I totally agree. Yeah. And I know David and the work they do is really cool. So I think that's an interesting area to see how that takes off with scouts because again, you then come back to, well, is this going to take my job? Is it going to, you know, influence it? I think we've generally found that AI, if you want to call it AI, is enriching. What the scout can do, you know, should help them be more precise. But yeah, I think that's a really interesting area to see it evolving.
- [47:03]
Michael Bailey
Just one more, if I may, Paul. I remember we had a chat and, and you said when you kind of started out on your career journey, if you like, that there were things that you were seeing that have taken almost 10 years to come to pass. Because when they were there in front of you, the industry wasn't ready for them. As you sit here now, what are the things that you think the football industry isn't ready for until maybe another 10 years time?
- [47:25]
Paul Power
Yeah, that's a great question. I think 10 years ago, so we'd built pressure models and looking at defensive balance and overloads and underloads, but again, being Able to generate that at scale and for teams to use that was. Was really difficult. I think the, the big thing is going to be about this body pose. So if you look at baseball, where I take a lot of my inspiration from on this, there's a great book, the MVP Generator or something that it's called, where it's looking at the pictures. They have these super slow cameras and they can basically see where the fingers are placed to be able to generate spin, which has come across into cricket as well. And, you know, they're able to take a guy who's average and make these guys the best. And it's like, how do we do that in football? You know, because it's, again, it's more complex because you can't really zoom into the player and capture everything. But for me, the idea of this body pose, about understanding stance and posture, you know, the ability to turn. Are you able to get down low? What does that mean? Do you prevent yourself from being explosive so you can then evaluate and train the person to move better? So I think Ajax did something similar with this, with the goalkeepers. Yeah, I think that's kind of where it's. It's going to lead to is it's more about talent development rather than talent recruitment, which I think is what data has really, you know, enabled clubs to improve in.
- [48:43]
Michael Bailey
That's fascinating. Well, absolutely brilliant chat. Thank you so much for coming in, Paul.
- [48:46]
Paul Power
No problem.
- [48:47]
Michael Bailey
Absolute pleasure. That is all we've got time for, sadly. A huge thank you as well to Mark and to Michael as well as producer Mike. Please do keep all the messages coming in. We do read them all and we will see you all next week. Until then, please do enjoy the football.
- [49:04]
eBay Advertiser
The Athletic FC podcast Network. As we progress through the season, every fan knows that big wins are hard to come by and tough losses are even harder to explain. Accept. But you know what isn't hard to accept? Discover. Believe it or not, Discover is accepted at 99 of places that take credit cards nationwide. You heard that right, 99. So make a good call for your wallet and get Discover. Based on the February 2024 Nielsen report. Learn more at discover.com credit card.