WEBVTT - Moneyball, Soccer, and the Gap Between Analytics and the Real World 0:00:08.400 --> 0:00:19.840 Pushkin. Around twenty ten, Sarah Rudd was rooting for this 0:00:20.079 --> 0:00:23.639 British soccer team. It was a team called Blackpool. It 0:00:23.680 --> 0:00:26.880 had been her great grandfather's favorite team. She was really 0:00:26.920 --> 0:00:30.560 into it, but she started getting annoyed by this one player. 0:00:30.960 --> 0:00:32.400 His name was Charlie Adam. 0:00:32.720 --> 0:00:36.479 He just loved to take shots from really long distance 0:00:36.760 --> 0:00:40.080 and it drove me crazy. Every possession just kind of 0:00:40.200 --> 0:00:43.839 ended with him taking like these random shots and it's like, 0:00:43.920 --> 0:00:44.800 what a waste? 0:00:44.880 --> 0:00:45.440 What are you doing. 0:00:45.960 --> 0:00:48.879 Sarah was living in Seattle working for Microsoft as a 0:00:48.920 --> 0:00:53.760 software engineer, but she really wanted to get into sports analytics. Basically, 0:00:53.880 --> 0:00:56.880 she wanted to do moneyball, but for soccer. And then 0:00:57.000 --> 0:01:00.480 she discovered that this sports data company called stat DNA 0:01:00.760 --> 0:01:04.360 was hosting a contest, a contest where people could use 0:01:04.400 --> 0:01:10.200 the company's soccer tracking data to generate analytical insights. And 0:01:10.240 --> 0:01:13.399 she thought this could be my chance, This could be 0:01:13.440 --> 0:01:16.800 my chance to break into the industry, and also at 0:01:16.800 --> 0:01:20.120 the same time that that guy, Charlie Adam really should 0:01:20.160 --> 0:01:22.720 not be taking those long shots. 0:01:22.959 --> 0:01:27.440 I'm a big fan of searching for inspiration elsewhere, so 0:01:27.720 --> 0:01:29.959 I started just looking around and seeing like, well, what 0:01:30.040 --> 0:01:30.759 have other people done? 0:01:30.800 --> 0:01:34.679 In this space, and it turns out for soccer not much. 0:01:35.080 --> 0:01:38.040 But there was some really interesting papers in other sports, 0:01:38.080 --> 0:01:40.680 and the one that caught my eye was by a 0:01:40.720 --> 0:01:44.920 gentleman named Keith Goldner. He had done something using NFL 0:01:45.000 --> 0:01:48.200 data where he looked at kind of this, if the 0:01:48.240 --> 0:01:51.000 ball is on the thirty five yard line, it's second 0:01:51.080 --> 0:01:53.920 and ten with you know, X amount of time on. 0:01:53.920 --> 0:01:55.639 The clock, like how likely are you to score? 0:01:56.360 --> 0:01:59.320 And you know, I thought that was really interesting and 0:01:59.400 --> 0:02:01.480 I thought, well, isn't that kind of similar to Charlie 0:02:01.480 --> 0:02:04.640 Adam having the ball forty yards out? 0:02:04.880 --> 0:02:08.560 So Sarah took this methodology it's called a Markov chain 0:02:08.720 --> 0:02:12.120 just fyi, and she applied it to soccer. She built 0:02:12.160 --> 0:02:14.480 a model that let you look at a moment when 0:02:14.520 --> 0:02:17.280 a player had the ball in a given location on 0:02:17.400 --> 0:02:20.840 the field, and the model could evaluate what the player 0:02:20.960 --> 0:02:25.480 did at that moment. Specifically, the model let you ask, 0:02:25.960 --> 0:02:29.840 did the choice the player made passing forward, passing to 0:02:29.880 --> 0:02:34.960 the side, shooting whatever. Did this choice the player made 0:02:35.440 --> 0:02:39.760 increase or decrease the probability of the team scoring a 0:02:39.800 --> 0:02:41.480 goal on that possession? 0:02:41.960 --> 0:02:45.600 And yeah, it really hadn't been done in soccer before, 0:02:45.680 --> 0:02:49.799 which was kind of yeah, why it was kind of 0:02:49.840 --> 0:02:50.799 a revolutionary paper. 0:02:50.840 --> 0:02:52.000 I hate to use that term, but. 0:02:52.000 --> 0:02:54.440 Like I'm saying it listed, I'm saying that listed five 0:02:54.440 --> 0:02:56.320 point thirty eight had like a list of the ten 0:02:56.360 --> 0:02:59.040 big papers and figuring out this kind of thing, and 0:02:59.120 --> 0:03:02.000 yours was the first out on the list, so it 0:03:02.040 --> 0:03:04.760 was chronological to be fair, but still big first is 0:03:04.760 --> 0:03:05.280 a big deal. 0:03:05.520 --> 0:03:07.720 Yeah, and you know it was cool because I was 0:03:07.919 --> 0:03:09.480 kind of like, oh, well, I'm just doing what this 0:03:09.520 --> 0:03:12.120 guy in the other football is doing. 0:03:12.680 --> 0:03:13.920 But yeah, it was pretty impactful. 0:03:14.880 --> 0:03:17.280 And by the way, what did you find out about 0:03:17.360 --> 0:03:17.880 Charlie Adam? 0:03:18.240 --> 0:03:22.120 Yeah? Not good to shoot from my hearts out was 0:03:23.120 --> 0:03:26.920 what I found out, which was really like soul soothing 0:03:27.000 --> 0:03:28.600 for me, or it was like I was right. 0:03:34.200 --> 0:03:37.320 I'm Jacob Goldstein and this is what's your problem. Today 0:03:37.320 --> 0:03:40.320 we have the third and final episode in our series 0:03:40.360 --> 0:03:43.440 of interviews with people who are working at the frontiers 0:03:43.480 --> 0:03:47.440 of technology to help a lead athletes perform better. My 0:03:47.520 --> 0:03:50.720 guest today is Sarah Rudd. She's the co founder and 0:03:50.840 --> 0:03:55.160 CEO at Source Football, a soccer analytics company that works 0:03:55.160 --> 0:03:58.720 with professional soccer teams in the US, the UK, and Europe. 0:03:59.160 --> 0:04:02.600 There are different to frame the core problem that Sarah 0:04:02.680 --> 0:04:05.160 is trying to solve for today's show, I'm going to 0:04:05.200 --> 0:04:09.400 go with a problem that generalizes way beyond sports. The 0:04:09.480 --> 0:04:13.320 problem is this, how do you translate analytical insights into 0:04:13.400 --> 0:04:18.839 meaningful changes in the real world. In our conversation, Sarah 0:04:18.839 --> 0:04:21.599 and I talked about analytics and the challenge of getting 0:04:21.640 --> 0:04:24.720 people to change their behavior, change the way they make 0:04:24.800 --> 0:04:28.160 decisions based on analytics. But before we got to all that, 0:04:28.240 --> 0:04:30.679 we talked a little bit more about that first paper 0:04:30.760 --> 0:04:34.039 that Sarah wrote to prove that Charlie Adams should stop 0:04:34.040 --> 0:04:37.360 shooting from forty yards out. Sarah told me that paper 0:04:37.440 --> 0:04:39.880 also led to some less obvious insights. 0:04:41.360 --> 0:04:43.839 Yeah, I think like one of the first ones was 0:04:43.839 --> 0:04:47.120 that crossing from kind of like wide situations, which was 0:04:47.200 --> 0:04:50.480 real really still very popular. So if you think of 0:04:50.520 --> 0:04:52.920 like a player coming down the wings and then they're 0:04:52.960 --> 0:04:54.719 going to kick the ball high in the air and 0:04:54.720 --> 0:04:56.800 then hopefully a teammate is going to head it into 0:04:56.800 --> 0:04:57.360 the goal. 0:04:57.560 --> 0:05:00.000 That is also like pretty low value. 0:05:00.160 --> 0:05:03.719 Oh interesting, yeah, because you have like a really high 0:05:04.160 --> 0:05:06.320 chance of like turning the ball over from that. 0:05:06.920 --> 0:05:10.120 And was that finding contrary to the sort of conventional 0:05:10.160 --> 0:05:11.040 wisdom of the time. 0:05:11.360 --> 0:05:13.560 Yeah, at that time, it was at that time there 0:05:13.640 --> 0:05:18.520 was still a lot of teams using that tactic, and 0:05:18.560 --> 0:05:22.520 I think, you know, going back, like people were starting 0:05:22.520 --> 0:05:25.279 to just kind of intuitively be like, maybe there's a 0:05:25.320 --> 0:05:28.560 better way. So it's not like it kind of like 0:05:28.600 --> 0:05:32.200 broke soccer or football, but it just kind of I 0:05:32.240 --> 0:05:35.039 think reinforced some of the intuitions that people had where 0:05:35.040 --> 0:05:36.600 it's like, maybe there's a better way to do this. 0:05:36.880 --> 0:05:38.880 And just to be clear, in the time since then, 0:05:38.920 --> 0:05:42.320 in the whatever fifteen years or so since then, have 0:05:42.920 --> 0:05:45.520 analytics shown clearly that that's a bad call and has 0:05:45.600 --> 0:05:47.880 that style of play decreased as a result. 0:05:48.480 --> 0:05:50.640 Yeah, And you know, now it's actually seen as like 0:05:50.680 --> 0:05:53.159 a sign of like ooh, the offense is struggling, like 0:05:53.200 --> 0:05:56.680 we're resorting to this really low probability tactic. 0:05:57.240 --> 0:06:00.960 So okay, so this paper you write, you hired by 0:06:00.960 --> 0:06:04.960 this analytics company, and then the analytics company gets acquired 0:06:05.000 --> 0:06:08.640 by Arsenal, the famous London soccer team, and you wind 0:06:08.720 --> 0:06:13.479 up working at Arsenal for many years. And one of 0:06:13.520 --> 0:06:17.479 the things I've heard you talk about about that that time, 0:06:17.880 --> 0:06:20.640 and that clearly is a big important theme that goes 0:06:20.640 --> 0:06:23.640 beyond soccer is trying to figure out how to get 0:06:23.960 --> 0:06:27.920 the people who make real world decisions to actually listen 0:06:28.000 --> 0:06:30.599 to you, to you and the analytics people. 0:06:31.279 --> 0:06:34.839 Yeah, it's it's really hard because like if you you know, 0:06:34.920 --> 0:06:37.719 remove sports from it, just getting anybody to change their 0:06:37.880 --> 0:06:42.839 opinion based on facts or evidence is really really difficult. 0:06:42.880 --> 0:06:46.800 And now you're talking about people who are in incredibly 0:06:46.839 --> 0:06:50.799 stressful jobs where they can lose their job really based 0:06:50.800 --> 0:06:51.240 on like. 0:06:51.320 --> 0:06:54.080 Some random occurrence that happens over the weekend. 0:06:54.240 --> 0:06:57.240 Right, Like, right, the result doesn't go other way, you're fired, 0:06:57.680 --> 0:06:58.400 Good luck to you. 0:06:58.400 --> 0:07:01.240 Which is not the way analytics works. Right, Like, the 0:07:01.240 --> 0:07:03.680 way analytics works is you need a large end, You 0:07:03.760 --> 0:07:05.880 need to do the thing one hundred times, and then 0:07:06.320 --> 0:07:08.600 sixty times it'll go the way the analytics says it. 0:07:08.640 --> 0:07:11.560 Forty times it won't. That's just the nature of the 0:07:11.600 --> 0:07:13.000 probabilistic world we live in. 0:07:13.360 --> 0:07:16.040 Yeah, and you know, like analytics is also all about 0:07:16.120 --> 0:07:19.280 separating process from outcomes, and yet these decisions are still 0:07:19.320 --> 0:07:22.280 being made on outcomes. So I have a lot of 0:07:22.280 --> 0:07:25.120 empathy for kind of people who are in this situation. 0:07:25.280 --> 0:07:27.720 And a coach will get fired more likely if they 0:07:27.840 --> 0:07:30.720 do the thing that the analytics says that's contrary to 0:07:30.760 --> 0:07:33.040 the sort of conventional wisdom in the sport. Right. It's 0:07:33.040 --> 0:07:36.520 like why in American football coaches for a long time 0:07:36.560 --> 0:07:39.200 didn't go for it enough on fourth down, right when 0:07:39.280 --> 0:07:42.280 the analytic clearly said they should, But everybody would get 0:07:42.280 --> 0:07:43.760 pissed at them if they went for it and didn't 0:07:43.760 --> 0:07:47.920 get it. Now that's change right because of analytics. Interestingly, although, 0:07:48.040 --> 0:07:49.880 was it this last Super Bowl where the coach kept 0:07:49.880 --> 0:07:51.360 going for it on fourth down and not getting it 0:07:51.400 --> 0:07:53.640 and people are like, he did it too much? I 0:07:53.680 --> 0:07:55.960 was like, no, he didn't. Just because it didn't work 0:07:56.000 --> 0:07:57.840 out doesn't mean it was the wrong decision. 0:07:58.480 --> 0:08:00.280 Yeah, And so you know, this just kind of goes 0:08:00.320 --> 0:08:02.720 back to like you need to have kind of top 0:08:02.760 --> 0:08:05.560 top decision makers saying like, yeah, it's okay if you're 0:08:05.560 --> 0:08:08.080 going to do something that looks a little bit unconventional. 0:08:08.240 --> 0:08:10.120 Yeah, we believe in it, we trust it. 0:08:11.240 --> 0:08:13.360 So okay. So we've talked about why it's hard. Did 0:08:13.400 --> 0:08:15.920 you figure anything out about how to get people to 0:08:16.000 --> 0:08:16.840 change their decision? 0:08:16.960 --> 0:08:19.640 I would say the best thing would be building trust 0:08:19.800 --> 0:08:22.840 through a common language, And so for us, the common 0:08:22.880 --> 0:08:26.720 language was video, so we could build a model but 0:08:26.880 --> 0:08:29.760 like until we could show them the model outputs on 0:08:30.000 --> 0:08:33.439 video and walk them through it and say, this is 0:08:33.480 --> 0:08:36.319 what the model sees, this is what the model predicts. 0:08:36.679 --> 0:08:39.040 What do you think, like, walk me through it. And 0:08:39.120 --> 0:08:42.360 so then it became kind of like these collaborative, iterative 0:08:42.559 --> 0:08:43.800 model building processes. 0:08:44.120 --> 0:08:46.880 It's sort of a version of having an idea and 0:08:46.880 --> 0:08:49.959 then convincing your boss that that idea is actually their idea, 0:08:50.440 --> 0:08:52.840 and once they think it's their idea, then they'll do it. 0:08:52.840 --> 0:08:55.959 It sounds like that, like maybe a little bit less cynical, 0:08:56.040 --> 0:08:56.760 but yeah. 0:08:56.640 --> 0:09:00.199 Yeah, yeah. So so eventually you left Arsenal and you 0:09:00.280 --> 0:09:04.160 started source Football, this company that you're running. Now, what 0:09:04.160 --> 0:09:05.400 what led you to make that leap? 0:09:05.840 --> 0:09:08.840 So, you know, part of starting this company was that, 0:09:09.160 --> 0:09:11.480 you know, I wanted to learn and to experience as 0:09:11.520 --> 0:09:14.960 much as possible. But the other part is that we 0:09:15.040 --> 0:09:19.040 felt like there's a huge need for clubs to get 0:09:19.040 --> 0:09:22.560 help in terms of getting started on this analytics journey 0:09:22.559 --> 0:09:24.880 where you know, a lot of them just don't even 0:09:25.200 --> 0:09:25.920 know where to begin. 0:09:26.000 --> 0:09:27.680 They don't know what's good what's bad. 0:09:28.559 --> 0:09:31.840 So our kind of main value add is coming in 0:09:32.240 --> 0:09:34.960 helping them get set up, helping them get started and 0:09:35.000 --> 0:09:38.920 then once that happens, doing all of the hard work, 0:09:39.040 --> 0:09:42.959 hard thinking that's impossible to get done within a football club. 0:09:43.920 --> 0:09:46.960 If anyone has ever been in a training ground, Like 0:09:47.080 --> 0:09:50.960 I'm sure it's the same across all sports, but they're chaotic, 0:09:51.080 --> 0:09:54.360 they're noisy, they're loud, Like they're not good places to 0:09:54.440 --> 0:09:59.120 do deep testosterone. Yeah, a lot of a lot of 0:09:59.120 --> 0:10:04.240 really loud music, like humping. You know, like if your 0:10:04.240 --> 0:10:06.679 office is anywhere near the gym, like forget it. Like 0:10:06.920 --> 0:10:08.760 you got to wait until everybody goes home before you 0:10:08.760 --> 0:10:11.400 can actually like have a clear thought. So you know, 0:10:11.440 --> 0:10:14.640 that was that was one of the things that we realized, 0:10:14.760 --> 0:10:16.720 is like you can't work on solving these really hard 0:10:16.720 --> 0:10:20.240 problems and like football is really hard to analyze in 0:10:20.320 --> 0:10:22.360 terms of analytics, and like you just you can't do 0:10:22.400 --> 0:10:23.480 it within a club. 0:10:24.160 --> 0:10:25.400 So that's why we want it to be a little 0:10:25.440 --> 0:10:26.000 bit outside. 0:10:26.000 --> 0:10:28.319 And then kind of our our long term vision is 0:10:28.360 --> 0:10:32.200 actually like you know, we're using this consulting business to 0:10:32.280 --> 0:10:36.360 kind of fund the development of our intelligence platform, and 0:10:36.400 --> 0:10:38.880 then the ideas eventually to look for like a set 0:10:38.880 --> 0:10:42.600 of investors or maybe an ownership group and you know, 0:10:42.720 --> 0:10:44.920 take control of a club and run it in kind 0:10:44.920 --> 0:10:47.360 of like the modern progressive way that we think they 0:10:47.360 --> 0:10:48.000 should be run. 0:10:48.679 --> 0:10:52.240 So what you really want to do is basically buy 0:10:52.240 --> 0:10:54.720 a soccer team and run it smarter than everybody else. 0:10:54.800 --> 0:10:57.960 Is that what you're telling me? Yeah, of course, tell 0:10:58.000 --> 0:10:59.800 me more about the big gream. We'll do the pieces. 0:10:59.840 --> 0:11:02.559 But I'm curious, Like that's a big, audacious dream and 0:11:02.640 --> 0:11:03.640 it's fun tell me about it. 0:11:03.840 --> 0:11:05.160 Yeah, I mean, I think you know, one of the 0:11:05.160 --> 0:11:08.719 things that probably everybody in analytics has experienced is that 0:11:08.960 --> 0:11:11.920 unless you're kind of like the key decision maker, it's 0:11:12.000 --> 0:11:16.080 always going to be kind of difficult to influence decisions 0:11:16.880 --> 0:11:19.680 because there's always going to be somebody that has, you know, 0:11:19.800 --> 0:11:23.480 kind of their own perspective and everything. And you know, 0:11:23.520 --> 0:11:25.560 at Arsenal, we were really lucky because we had like 0:11:25.600 --> 0:11:27.599 a lot of resources. But what I see at a 0:11:27.640 --> 0:11:30.760 lot of clubs is that you know, they only hire 0:11:30.800 --> 0:11:33.600 one or two people. The work that they do is good, 0:11:33.720 --> 0:11:36.200 but like there's a limit to what two human beings 0:11:36.240 --> 0:11:39.040 can do, and so like a lot of what they 0:11:39.040 --> 0:11:42.840 produce doesn't necessarily like match the gut, and so it's 0:11:42.840 --> 0:11:46.600 hard to get this buy in. And so then they 0:11:46.640 --> 0:11:49.120 just say like, well, okay, thanks for that. I'm only 0:11:49.120 --> 0:11:50.640 going to listen to you if I agree with it. 0:11:50.960 --> 0:11:52.600 If I disagree with it, I'm going to go with 0:11:52.679 --> 0:11:55.760 my gut, and so it just becomes difficult. 0:11:55.800 --> 0:11:58.160 But I think also, like there's so many issues within 0:11:58.200 --> 0:12:02.960 football clubs that go beyond just analytics or you know, 0:12:03.080 --> 0:12:05.120 making the right decision on players. I mean, there's so 0:12:05.200 --> 0:12:08.880 much in terms of like building good, good cultures, you know, 0:12:09.000 --> 0:12:12.640 kind of making sure that everything is kind of set 0:12:12.720 --> 0:12:15.360 up in like a professional way, because if you think 0:12:15.400 --> 0:12:17.240 about it, clubs are kind of run by people who've 0:12:17.240 --> 0:12:18.480 never worked outside of football. 0:12:19.679 --> 0:12:20.719 But I think there's a lot of. 0:12:20.720 --> 0:12:24.520 Lessons that you can take from working outside of sports 0:12:24.840 --> 0:12:28.280 bring them into football, and we've seen it in every 0:12:28.320 --> 0:12:30.040 other major American sport. 0:12:30.440 --> 0:12:33.320 Not to be like super ruthless. 0:12:32.880 --> 0:12:35.160 And say like we got to maximize profit, maximize profit, 0:12:35.920 --> 0:12:39.559 because I think like they are these weird social institutions 0:12:39.600 --> 0:12:42.120 that have a lot of meaning to a lot of people, 0:12:42.160 --> 0:12:44.360 and so you obviously want to respect that. And you 0:12:44.400 --> 0:12:48.000 also have these amazing platforms to bring positive change into 0:12:48.040 --> 0:12:49.760 the world, and so you want to take advantage of 0:12:49.760 --> 0:12:50.280 that as well. 0:12:50.320 --> 0:12:51.760 But like, there certainly can be a. 0:12:51.760 --> 0:12:55.760 Lot more kind of professionalism in them than what we 0:12:55.840 --> 0:12:57.200 experience at a lot of places. 0:12:58.120 --> 0:13:00.480 So, now you have this company and you want to 0:13:00.559 --> 0:13:04.280 use it eventually to take over the world. Correct, before 0:13:04.440 --> 0:13:06.840 you use your company to take over the world, Like, 0:13:06.920 --> 0:13:09.600 what are the what are the services you're selling to 0:13:10.120 --> 0:13:11.160 teams to clubs? 0:13:11.600 --> 0:13:16.599 Yeah, I mean, so clubs are in really different situations. 0:13:16.679 --> 0:13:19.160 And then also, like you know, it's not like the 0:13:19.320 --> 0:13:22.600 US where like every single major League baseball team is 0:13:22.640 --> 0:13:27.800 like loaded with cash. Typically in European leagues, like you're 0:13:27.800 --> 0:13:30.400 going to have several divisions, so like you're talking about 0:13:30.440 --> 0:13:34.520 anywhere from like a single a baseball team to like 0:13:34.760 --> 0:13:39.280 a Major League baseball team, Like there's huge difference in resources. 0:13:38.840 --> 0:13:42.400 So like orders of magnitude in terms of revenue, how 0:13:42.480 --> 0:13:46.320 much presumably they want to spend on your services, et cetera. 0:13:46.720 --> 0:13:49.200 Yeah exactly, And so like the main area where we 0:13:49.200 --> 0:13:53.200 can help them is really recruitment, so helping them kind 0:13:53.200 --> 0:13:56.839 of find the best players. So unlike American sports where 0:13:56.840 --> 0:14:00.320 you kind of like trade players and draft them, everything 0:14:00.360 --> 0:14:02.280 here is kind of like an open market where you 0:14:02.280 --> 0:14:04.400 can buy and sell the contracts of these players. 0:14:04.440 --> 0:14:06.480 And so if you're a small team, and you have 0:14:06.520 --> 0:14:07.439 a really really. 0:14:07.559 --> 0:14:10.360 Good player, like you can just sell him keep that cash, 0:14:10.440 --> 0:14:12.760 or you can invest it back into your club. 0:14:13.160 --> 0:14:16.960 Huh. This is this is really echoes of Moneyball. I mean, 0:14:16.960 --> 0:14:18.480 I'm sure I don't know if you're tired of hearing 0:14:18.480 --> 0:14:20.680 about Moneyball. That book came about twenty years ago, but 0:14:20.800 --> 0:14:23.320 like that was the basic idea there, Right, it was 0:14:24.640 --> 0:14:27.920 better scouting essentially, right, Like you had these scouts who 0:14:27.920 --> 0:14:29.560 were sort of using their guts and these kind of 0:14:29.560 --> 0:14:33.520 conventional wisdom heuristics, and then you had what I'm sure 0:14:33.520 --> 0:14:37.480 now seem like primitive analytics coming in and basically doing 0:14:37.480 --> 0:14:41.600 a better job of predicting player success, right at some level. 0:14:41.760 --> 0:14:42.560 Yeah, exactly. 0:14:42.600 --> 0:14:44.440 And so you know, I hate to say that, like 0:14:45.040 --> 0:14:47.680 soccer is twenty years behind baseball, but like in a lot. 0:14:47.520 --> 0:14:50.040 Of ways we are. And so like this is still 0:14:50.080 --> 0:14:52.000 like the main area of. 0:14:52.000 --> 0:14:54.640 Value add for a lot of clubs because you're competing 0:14:54.680 --> 0:14:58.760 in this open market with people with varying amounts of knowledge. 0:14:58.840 --> 0:15:02.600 So sometimes that knowledge is just hey, like we had 0:15:02.640 --> 0:15:05.080 scouts at that game, We've seen this player once, like 0:15:05.720 --> 0:15:08.320 this is our opinion of him. And then you have 0:15:08.520 --> 0:15:10.600 the more sophisticated teams that are like, Okay, well, we 0:15:10.640 --> 0:15:14.280 have a database of you know, six hundred thousand matches 0:15:14.640 --> 0:15:17.960 in the world. We know the fifty best players in 0:15:18.040 --> 0:15:20.920 every league at every position, we know how much they're worth. 0:15:21.920 --> 0:15:24.360 So we help teams kind of go from like the 0:15:24.400 --> 0:15:27.040 former to the latter and just be a little bit 0:15:27.080 --> 0:15:29.640 more sophisticated in terms of the amount of information that 0:15:29.680 --> 0:15:29.960 they have. 0:15:30.160 --> 0:15:32.080 I mean, it's sort of crass to put it this way, 0:15:32.120 --> 0:15:35.880 but it's really like pricing assets right, like the players. 0:15:35.960 --> 0:15:38.640 And again I realized this is kind of dehumanizing. I'm sorry, 0:15:38.680 --> 0:15:43.040 but it is analogous to people valuing a stock or 0:15:43.160 --> 0:15:48.360 valuing anything they might buy. Right, And the better you 0:15:48.400 --> 0:15:52.640 can model the value of the asset, the better you're 0:15:52.640 --> 0:15:55.160 going to be at finding mispriced assets. Right. You want 0:15:55.200 --> 0:15:59.400 to you want to find bargains. You want to get 0:15:59.400 --> 0:16:02.440 the most you can and uh for your dollar. 0:16:02.960 --> 0:16:06.080 Yeah, absolutely, And so you know, there's a lot of 0:16:06.800 --> 0:16:09.320 techniques that we can take from other industries because it's 0:16:09.360 --> 0:16:11.680 not that different from it. Where it gets hard is 0:16:11.680 --> 0:16:14.160 that you have to have those assets work well together. 0:16:14.560 --> 0:16:16.000 They are in fact human beings. 0:16:16.640 --> 0:16:18.520 Yes, that is you know the. 0:16:18.880 --> 0:16:20.720 Kind of joke where it's like, well, the problem with 0:16:20.720 --> 0:16:23.080 football clubs is that it's full of human beings. 0:16:23.760 --> 0:16:27.040 So let's talk about what you can model, Like, what 0:16:27.120 --> 0:16:29.480 are you good at modeling in this context, like in 0:16:29.560 --> 0:16:33.280 helping in scouting, essentially in helping clubs value, you know, 0:16:33.360 --> 0:16:34.560 players they might acquire. 0:16:35.120 --> 0:16:38.360 Yeah, so we're pretty good at modeling everything that happens 0:16:38.400 --> 0:16:42.640 when a player has the ball, and unfortunately in football, 0:16:42.760 --> 0:16:45.800 that's like a very very small portion of the gaming. 0:16:46.240 --> 0:16:48.320 Yeah. And so there was this really cool. 0:16:48.040 --> 0:16:51.320 Development in the last four or five years where a 0:16:51.400 --> 0:16:54.720 number of companies have come out with a data source 0:16:55.080 --> 0:16:59.800 that's basically taking the video feeds from TV and turning 0:17:00.280 --> 0:17:04.760 into tracking data. So basically they're able to track the 0:17:04.800 --> 0:17:06.959 location of every player that's on screen. 0:17:07.760 --> 0:17:10.520 So obviously you don't know all the players, and. 0:17:10.440 --> 0:17:12.960 Then you know, what we've learned is that really what's 0:17:13.000 --> 0:17:17.400 going on off screen tends to not be as relevant 0:17:17.440 --> 0:17:21.320 and the location of those players isn't as relevant. And 0:17:21.359 --> 0:17:23.040 so now all of a sudden, you can start doing 0:17:23.119 --> 0:17:26.879 modeling on what people are doing off the ball and 0:17:27.040 --> 0:17:30.159 use it for recruitment. Because prior to this, similar to 0:17:30.200 --> 0:17:32.800 the NBA, there were cameras in all the stadiums and 0:17:32.840 --> 0:17:36.240 you would get this full data set, but only for 0:17:37.040 --> 0:17:38.919 your league, and so you couldn't really use it for 0:17:38.960 --> 0:17:39.440 a recruitment. 0:17:39.440 --> 0:17:41.280 And so this has been like a really big change. 0:17:41.320 --> 0:17:45.160 So you know, we can get much better views into 0:17:45.320 --> 0:17:48.440 what is a player doing physically, what are they doing 0:17:48.480 --> 0:17:52.399 defensively in terms of cutting off passing lanes, things like that. 0:17:52.920 --> 0:17:55.280 There's still a lot we're not good at with that, 0:17:56.040 --> 0:17:58.920 but what it's also done is allowed us to say, well, 0:17:58.920 --> 0:18:01.080 if we're going to move a player from this league 0:18:01.080 --> 0:18:04.760 to that league, how different is it physically? How do 0:18:04.800 --> 0:18:07.000 we think they're going to adapt? And so this has 0:18:07.040 --> 0:18:09.119 really opened up a lot of things. So if we 0:18:09.160 --> 0:18:12.840 want to talk about like risk adjusted pricing of assets, 0:18:13.359 --> 0:18:16.200 now we're starting to be able to quantify a little 0:18:16.240 --> 0:18:19.160 bit like what's the risk of bringing in somebody from 0:18:19.240 --> 0:18:23.240 a really really different environment into this one versus one 0:18:23.280 --> 0:18:24.880 that's more similar, Or. 0:18:25.440 --> 0:18:29.199 So you should apply a larger discount when you're bringing 0:18:29.200 --> 0:18:32.000 a player from a very different league, presumably because you're 0:18:32.040 --> 0:18:32.840 taking more risk. 0:18:33.240 --> 0:18:36.520 Yes, yes, and if not, then it's not an undervalued 0:18:36.560 --> 0:18:38.440 asset and maybe walk away. 0:18:39.200 --> 0:18:44.360 So tell me more, like, is the output you have 0:18:45.119 --> 0:18:47.439 just here is what you should pay for each of 0:18:47.480 --> 0:18:51.000 these players, Like you have whatever ten thousand players or something, 0:18:51.040 --> 0:18:53.119 and you put a dollar value on each one, or 0:18:53.240 --> 0:18:54.520 like what's the output? 0:18:55.040 --> 0:18:56.959 Yeah, I mean I wish it was that simple. 0:18:57.000 --> 0:18:59.960 We're still like pretty far from putting it all together 0:19:00.040 --> 0:19:03.199 there and saying like, by this player at this price. 0:19:04.200 --> 0:19:07.000 You know, a lot of the difficulties is that there's 0:19:07.000 --> 0:19:10.200 no good data set on pricing information. 0:19:10.000 --> 0:19:11.960 So then what do you just have a relative like 0:19:12.040 --> 0:19:13.040 kind of a stack rank. 0:19:14.000 --> 0:19:17.040 I mean, there are like estimates, so in the in 0:19:17.080 --> 0:19:20.080 the media, they'll say like, oh, this guy went for 0:19:20.440 --> 0:19:23.160 ten million, but then a different media source will say 0:19:23.160 --> 0:19:27.000 this guy went for fifteen million, because like the selling 0:19:27.040 --> 0:19:29.879 club wants to report a higher price and the buying 0:19:29.920 --> 0:19:31.920 club wants for lower price. 0:19:31.720 --> 0:19:33.560 So there's real no truth. 0:19:33.640 --> 0:19:39.479 And then salaries for players are not public for most leagues. 0:19:39.800 --> 0:19:42.919 So okay, so that's a very important variable that you 0:19:43.000 --> 0:19:45.399 don't really have access to. That's a problem. So what 0:19:45.800 --> 0:19:47.360 is the output of your model? Then? 0:19:47.680 --> 0:19:49.639 Yeah, I mean, so we we kind of do like 0:19:49.680 --> 0:19:53.719 a stack ranking and like a you know, projected like 0:19:53.800 --> 0:19:56.080 what do we think this player would do in this situation? 0:19:56.240 --> 0:19:58.399 But like the error bars on these things are like 0:19:58.640 --> 0:19:59.359 are pretty big. 0:19:59.480 --> 0:20:03.000 So we're still kind of in the like subjective realm 0:20:03.080 --> 0:20:06.040 of like based on these factors, we think this or 0:20:06.080 --> 0:20:08.159 we think that, and then you know a lot of 0:20:08.200 --> 0:20:12.760