{"id":17734,"date":"2025-07-24T05:01:11","date_gmt":"2025-07-24T05:01:11","guid":{"rendered":"https:\/\/www.newsbeep.com\/au\/17734\/"},"modified":"2025-07-24T05:01:11","modified_gmt":"2025-07-24T05:01:11","slug":"football-architects-how-the-sports-data-pioneers-convinced-the-world-to-take-notice","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/au\/17734\/","title":{"rendered":"Football Architects: How the sport\u2019s data pioneers convinced the world to take notice"},"content":{"rendered":"

This is the fourth of a six-part series looking at figures who have played a pivotal role in a modern football success story. The first piece, on the rebuilding of Ajax, can be found here<\/a>. Part two, on\u00a0Belgium becoming No 1 in the FIFA Rankings is here<\/a>. And part three, on the rise of Croatian football is here<\/a>.\u00a0<\/p>\n

Each article comes with a related podcast, which can be found\u00a0here on The Athletic FC Tactics Podcast feed<\/a>.<\/p>\n

\u201cAll this data analysis in football is bulls**t, isn\u2019t it?\u201d<\/p>\n

Ian Graham had been hired to assist Spurs\u2019 recruitment team, but his first meeting with Michael Edwards, his boss at Tottenham Hotspur, and future sporting director at Liverpool, was not going well.<\/p>\n

\u201cIt was long and aggressive,\u201d Graham explains. \u201cI called my boss (at Decision Technology \u2014 an analytics company) up after the meeting and said I thought Spurs are going to drop us, because they clearly think we\u2019re idiots. But he wanted to find out the answer. The way he asked them, I thought he was trying to make me look stupid, but he was honestly trying to gauge the quality of what we were doing.\u201d<\/p>\n

Working under Edwards and Jurgen Klopp at Liverpool, Graham\u2019s work helped turn the club into Premier League and European champions, transforming the squad in the process.<\/p>\n

But his first experience was typical of many of the first data pioneers of English football. They have now, however, broken through \u2014 their work becoming mainstream across recruitment, training methodology, and even the sport\u2019s dialect.<\/p>\n

\u201cWhen I started, realistically the brief was: \u2018We\u2019ve got lots of data, we think there\u2019s some value in it, can you find something?\u2019\u201d says Sam Green, a former data analyst at Opta who went on to work at Aston Villa. \u201cThat was it. It wasn\u2019t like we wanted to change football.\u201d<\/p>\n

\u201cYou\u2019re seeing fans having online arguments about expected goals (xG) and player age,\u201d adds Graham. \u201cBut at least they\u2019re arguments about things that make sense \u2014 things that clubs actually worry about, rather than the old, \u2018Who\u2019s best in the dressing room?\u2019 stuff.\u201d<\/p>\n

Generally, the birth of football\u2019s data revolution is dated back to the mid-2000s, although companies such as Opta and Prozone had been collecting data since the late 1990s. Back then, even if the ideas were sound, the battle its proponents faced was for people to believe in it.<\/p>\n

Part of the issue was the existing data\u2019s limitations. Baseball, the poster-child for the analytics movement, had a dataset going back decades \u2014 in contrast, when Graham began at Spurs, he had only two seasons to work with, with many foreign leagues completely inaccessible. While in baseball, statistics recording the outcome of every pitch told analysts what actually happened \u2014 high inside, low outside, where hit, how out \u2014 football\u2019s existing data did not necessarily do the same. What inherent value was there in knowing a player completed 19 passes?<\/p>\n

\u201cFootball is one of the most complex sports to analyse, just because of the sheer number of people involved,\u201d explains Green, one of the key players in the invention of xG. \u201cYou need to worry about the dynamics of how those 22 players are interacting, and it just makes it much harder. After calculating the simple, event-based data, you don\u2019t hit a wall as such but it becomes more and more expensive, and the number of people with access just gets smaller and far less public.\u201d<\/p>\n

\u201cWhat is a pass, what is an aerial challenge, what is a clearance?\u201d asks Sarah Rudd, one of the first full-time in-house data analysts at Arsenal. \u201cIt\u2019s actually quite difficult. Was that player trying to cross or shoot? There\u2019s a lot of ambiguity around these things, so getting agreement across data collectors is tricky. And that creates this whole mess where we can\u2019t even really define a pass or a shot, and now are trying to build complex models on top of it.\u201d<\/p>\n

Many analysts in the 2000s found themselves tarred by the reputation of Charles Reep \u2014 a visionary in many ways with his painstaking, handwritten analysis into play patterns in the 1950s \u2014 but who was ultimately widely criticised for his conclusions, preaching long ball football over all other styles. His ideas were later propagated by the likes of Charles Hughes in the 1980s, arguably setting back English football stylistically by decades. In his book, How to Win the Premier League, Graham describes former Liverpool head coach Brendan Rodgers raising this example.<\/p>\n

But by this point, however, Graham had a supporter in Edwards, who had begun to see the value in his work.<\/p>\n

\"\"<\/p>\n

Michael Edwards (left), with Jurgen Klopp in 2019 (John Powell\/Liverpool FC via Getty Images)<\/p>\n

\u201cPart of the reason that Michael was initially sceptical was that he started off as a Prozone analyst, where he was employed to produce statistical reports for the Portsmouth coaches,\u201d Graham explains. \u201cBut he knew Harry Redknapp and his assistants would have thrown him out the room, because the report was kind of useless \u2014 showing number of sprints and total distance. Michael was a youth player; he knew from his football intuition that it didn\u2019t tell you anything about the game.<\/p>\n

\u201cSo part of his scepticism was because he had seen it done badly before. We had a very early version of expected threat (the chance of scoring based on the ball\u2019s current location) and he started tearing down all of the assumptions in it. \u2018Sometimes the backwards pass is the best option\u2019, he would say, which we had to admit we could not see in the data.<\/p>\n

\u201cLuka Modric was the player we disagreed on most (when both were together at Spurs). We said he was an above-average player but not top 10 in the Premier League. Michael insisted he was by far the best player in our squad, and he was right, because his skills didn\u2019t come across very well in the event data.\u201d<\/p>\n

It speaks to a near-universal experience faced by these early analysts. Despite being typecast as \u2018numbers people\u2019, at times, their most important skill was a social one \u2014 possessing the ability to communicate their work within a club in order to display its value.<\/p>\n

\u201cSocial skills may be pushing it a little bit; we\u2019re all introverts \u2014 that\u2019s your standard personality type,\u201d says Graham. \u201cBut you have to communicate. Richard Feynman, the Nobel Prize-winning theoretical physicist, is a great example. He works on complex stuff, quantum mechanics, but has this quote that if you can\u2019t explain your theory in five minutes to somebody serving you a beer at your local bar, you don\u2019t understand that theory.<\/p>\n

\u201cA lot of football club staff are not data experts \u2014 they shouldn\u2019t be data experts, they have too many other calls on their time and haven\u2019t had the luxury of spending three years on a maths degree. But they\u2019re usually intelligent people, enthusiastic, and it\u2019s your job as the data person to explain in clear football language how your model works.\u201d<\/p>\n

\"\"<\/p>\n

Luka Modric\u2019s event data did not do justice to his impact on the pitch (Ian Walton\/Getty Images)<\/p>\n

Over at Arsenal, Rudd worked for an early analytics company named StatDNA<\/a>, who were acquired by the Premier League club in 2012 to improve their player recruitment. She worked closely alongside the team\u2019s management, at first Arsene Wenger and his coaching staff, for 10 years.<\/p>\n

\u201cWe always relied on video as a communication tool,\u201d she explains. \u201cIt\u2019s not to say that everything had to pass the eye-test, but if you\u2019re trying to claim something was a good action, and it didn\u2019t look like one, people are going to push back. So you need to be able to explain the knock-on effects, and present it in a way that everyone can feel comfortable with.<\/p>\n

\u201cOne good example would be when we worked on what appeared to be a simple project for the Arsenal academy, where they wanted to measure switches of play. But as we sat down with the coaching staff \u2014 I think Steve Bould was there \u2014 he would say: \u2018Yes, it\u2019s a switch, but that\u2019s not quite right\u2019. And it turned out that what they actually wanted was switching the ball and creating an overload. So we were seeing that the player executing the switch did the right thing, but the supporting player did not. So we changed our definition of success in the model, and it was good coaching feedback for them that their messaging needed to change.\u201d<\/p>\n

But sometimes, analysts still needed to know when to pick their battles. Statsbomb founder Ted Knutson worked at Brentford for several years, where his work was highly valued, but his experiences at other clubs sometimes left him reliant on demonstrative results.<\/p>\n

\u201cIt was a lipstick-on-a-pig situation with some old-school directors of football,\u201d he says. \u201cYou can add the lipstick, but it\u2019s only going to make them angry if they\u2019re not open to this stuff. There was pushback for the first five years consistently \u2014 it still even happened last year with a coach. But, eventually, owners stopped allowing their football people to do this, because they saw that the best teams in the world were using this data \u2014 they knew they had to get on board.\u201d<\/p>\n

Once they gained a foothold, data analysts increasingly became able to point to their track record as evidence for their worth. At Brentford, for example, data was a key part of the club\u2019s philosophy. Sitting in League One, lacking the resources to buy top players, they instead gradually exchanged their way up the leagues, generating returns and reinvesting, trading the proverbial red paperclip for a house.<\/p>\n

\u201cIn my final years at Arsenal, there was a core football intelligence group that worked really well together,\u201d says Rudd. \u201cThe attitude would be focused around mitigating the risk of any signings between all these different information sources, whether they were data or more traditional scouting. If you look at Arsenal\u2019s transfer history during that time, from 2020 onwards, the ratio of hits to misses changed dramatically.<\/p>\n

\u201cFor example, we did some work around really bespoke, subjective data collection around defending. That was a massive analytics black hole, and it still is for most clubs around the world. While event data usually collects tackles and duels, that\u2019s just players being aggressive, but the art of defending is around positioning and anticipating play \u2014 not interventions.<\/p>\n

But we designed a way to collect our own information around this gap. It was a program to basically evaluate when players are doing the right thing and the wrong thing, incredibly sophisticated and labour intensive. You could simplistically think about it in terms of penalising a winger for not tracking back, and allowing a threat to develop behind him, rather than punishing the full-back who is left one-on-one. It took five years to come into existence, and Arsenal\u2019s centre-back signings in recent years have a pretty good track record.\u201d<\/p>\n

\"\"<\/p>\n

Arsenal have used bespoke data to target suitable centre-backs in the past decade (Justin Setterfield\/Getty Images)<\/p>\n

The data revolution was not a linear progression, or even limited to only a handful of pioneers. From the mid-2000s onwards, there were numerous development trains running concurrently \u2014 the betting research of Tony Bloom and Matthew Benham, owners of Brighton & Hove Albion and Brentford respectively, in-house research teams at Arsenal and Liverpool, plus external consultants such as Opta and Statsbomb. Often, they worked in complete independence from each other.<\/p>\n

\u201cSometimes I\u2019d go to meetings with the commercial guys who were selling our work,\u201d says Green, then working at Opta. \u201cAnd when I did, it was often quite cagey. You would get the impression that clubs were into it, but no one would show their cards as to whether you were ahead of them. I remember going to Chelsea and definitely knowing they were doing something, but having no clue whether it was valuable.\u201d<\/p>\n

Knutson argues the history of football data could be loosely split into four waves:<\/p>\n

The gambling wave \u2014 led by Benham and Bloom, in which football data was first used for commercial purposes
\nThe in-house wave \u2014 where clubs such as Arsenal and Liverpool picked up on these possibilities, and began doing their own work behind the scenes
\nThe democratisation of data \u2014 in which companies including Opta and Statsbomb began to sell tools which clubs of varying wealth across the football pyramid were able to access and implement
\nThe birth of tracking data \u2014 where the likes of SkillCorner are able to add objective physical data to the existing numbers, going past some of the capabilities of traditional scouting<\/p>\n

Early on, Green\u2019s brief at Opta, in terms of research and development, was broadly experimental.<\/p>\n

\u201cObviously baseball was the precursor, so we knew there was something there,\u201d he says. \u201cSo it never felt futile; there were always people that were engaged. But there were some bits of work you did that went nowhere, and other bits that took off, and often they weren\u2019t what I would have valued. It is surprising what cuts through.\u201d<\/p>\n

Green is one of the central figures behind the development of xG. While Rudd was concurrently developing a similar model at Arsenal in private, Green\u2019s initial blog post on the subject has morphed into a statistic well-known enough to be used on Match of the Day. \u201cIt\u2019s like Newton and Leibniz independently describing calculus at the same time,\u201d jokes Rudd.<\/p>\n

\u201cIt\u2019s strange, because it didn\u2019t necessarily take off at the time, but subsequently has,\u201d Green says. \u201cIt\u2019s ultimately trying to describe the game \u2014 to codify what you think is actually happening. If you take a pot shot from 25 yards out, it\u2019s less likely to go in than a header from point-blank range.<\/p>\n

\u201cSo it was a big building block for me. You had to get this in place to understand the value of generating different shots and scoring goals, because that\u2019s the aggregation that everyone is trying to do in football. But the mainstream success of xG is relatively surprising.\u201d<\/p>\n

\"\"<\/p>\n

The BBC introduced xG to Match of the Day in August 2017 (Peter Dench\/Getty Images)<\/p>\n

But surprise is a key emotion when it comes to data analysis \u2014 challenging prevailing bias is when the discipline is at its most powerful. For example, ultimately, Graham was one of the figures who convinced Liverpool to move for Andy Robertson \u2014 but at the same time, Green was interested in recommending the Hull City left-back for Aston Villa.<\/p>\n

\u201cThere\u2019s a quote from (baseball statistics savant) Bill James which outlines 90 per cent test,\u201d says Green. \u201cIf you build a metric, and 100 per cent is what you expect, you\u2019ve not built anything of use. If only 50 per cent is what you expect, then your model is probably wrong. But if you expect 90 per cent, then the remaining data is really interesting. Often that\u2019s only one or two people.<\/p>\n

\u201cI remember Andy Robertson scoring really well in one, which surprised me, because of Hull City\u2019s relegation. Even though I should have been disassociating myself completely from his profile, it\u2019s still hard to be completely dispassionate about the situation and how it impacts the perception of a player.\u201d<\/p>\n

This information sometimes gets back to players or coaches with vested interests. For example, when evaluating your club\u2019s own squad, a technical director might have his past signings criticised. A player who commissions data insights into their value might be upset when it suggests they are not worthy of that lucrative contract.<\/p>\n

TransferMarkt managing director Thomas Lintz tells a story about Cristiano Ronaldo blocking the website on Instagram after being displeased with his latest value update.<\/p>\n

\u201cPeople like being able to control their environment, and so there are times when you give coaches or agents information and they don\u2019t like what they hear,\u201d says Knutson. \u201cI get it \u2014 you fight for whatever you can get. But we have to try and weigh it back to objectivity, and tell them to take it or leave it. There have been times when we\u2019ve cut team deals loose because people wanted us to change our reports. It cost us money, but we felt that the alternative lacked integrity.\u201d<\/p>\n

\"\"<\/p>\n

Andy Robertson\u2019s data in his time at Hull City was notably strong (Ian Horrocks\/Sunderland AFC via Getty Images)<\/p>\n

It is at these moments the overall structure becomes important. Graham is quick to point out his fortune at working at Liverpool under Klopp and Edwards, where clear processes were in place.<\/p>\n

\u201cThe importance of being honest with the manager is underplayed in recruitment,\u201d he explains. \u201cYou have to tell them that though you think Player X might be the best player, they are not perfect, and so here are their weaknesses \u2014 are you comfortable with them?<\/p>\n

\u201cThat was a great thing about Jurgen \u2014 he was willing to accept weakness in return for strength. He said that if a player had some super strengths, he could cover up their deficiencies. As a recruitment department, we could give our honest opinion, and then it avoids the coach being upset when they turn out not to be able to do X, Y, or Z.\u201d<\/p>\n

At Arsenal, the structure gradually improved. \u201cWe were so early to the game that a lot of people still didn\u2019t really understand what we were doing or how best to use us,\u201d explains Rudd. \u201cI think one thing we struggled with at the start was that in the era before sporting directors, you had one manager in charge of absolutely everything. So of course they wouldn\u2019t necessarily put a lot of thought into what the decision-making process should look like, which sometimes hindered us from being as impactful as we could.<\/p>\n

\u201cSome people think that the role of data in recruitment is just for generating lists or for filtering. I\u2019m very much of the opinion that data and analytics has its own voice and should have an independent evaluation of players, separate from scouts, so that you have these diverse perspectives and you can kind of cover each other\u2019s blind spots.\u201d<\/p>\n

Now, almost every Premier League club has a technical director or its equivalent, giving data its place within the club\u2019s management structure. Of course, there are still vast gaps between the leading clubs \u2014 generally considered as Manchester City, Liverpool and Arsenal \u2014 and those further down the table.<\/p>\n

But some new hires in wider footballing roles even possess an analytics background \u2014 Ben Knapper, Arsenal\u2019s former loans manager, and Norwich City\u2019s new sporting director, came through at StatsDNA alongside Rudd. It demonstrates the discipline\u2019s new-found reach.<\/p>\n

What is the next argument for them to win? One remaining frontier is whether they will begin to influence managers in game \u2014 suggesting substitutions or changes of system. Certain advancements in technology mean aspects of this are theoretically possible, even though the bar is so high to clear in practice.<\/p>\n

\u201cOne of the reasons why it\u2019s so hard is that there are multiple ways to win \u2014 it\u2019s not like baseball where there\u2019s an ideal and finite way to move players around the bases,\u201d explains Rudd. \u201cYou can really optimise that in a way you can\u2019t mirror in football, because whatever strategy you come up with, somebody is going to counterbalance that. Plus, 45 minutes is not a lot of data to make decisions on.\u201d<\/p>\n

She laughs at the irony of a data analyst making her next point. \u201cTo be honest, this is where the human brain just does a much better job at finding patterns. Pretty early on at Arsenal we got access to live tracking data, and the coaches wanted us to identify spaces we could exploit. But those spaces don\u2019t stay open for very long \u2014 so a human is just far better at saying: \u2018This person is positioned badly, or that player\u2019s reaction is a bit slow\u2019. With current tools, you could probably build a slightly better model, but I think this is just where the human brain really excels.\u201d<\/p>\n

Graham agrees. \u201cI don\u2019t think data has done much in terms of tactics yet, is the simple answer. You need a good tracking model to say something about tactics, and most teams don\u2019t have that \u2014 maybe just the big Premier League teams and a couple in Europe. It took us two years to build one at Liverpool and we had one of the best-resourced data science teams in football.<\/p>\n

\u201cBut for most teams, if data is informing tactics it probably shouldn\u2019t be unless you\u2019re sure you have a very good model. You have to convince the coach to use it as well, and that\u2019s a very high bar.\u201d<\/p>\n

\"\"<\/p>\n

Data is yet to much impact on in-game tactics (Matt McNulty\/Getty Images)<\/p>\n

Instead, some see the next frontier as taking place in training. Ninety per cent of Graham\u2019s role concerned recruitment and, though Rudd\u2019s focus on signings only comprised half her job, transfers were seen as the main priority. However, this means other areas possess low-hanging fruit.<\/p>\n

\u201cTraining data and improving player skills is going to be a big thing going forward,\u201d says Knutson. \u201cUsing equipment like high-speed camera and LiDAR radar can help improve ball-striking \u2014 that could have a big impact because better ball-striking has a very large impact in scoring goals. But it also has an impact on things like, for example, your centre-back playing the ball out to the wings consistently \u2014 and we know that\u2019s a super valuable pass that not every centre-back can make.\u201d<\/p>\n

Data analysts have had to fight to defend their role in football. Sometimes the numbers do not speak for themselves, but need somebody to speak up for them. For Graham, this is the only place where bias can be justified.<\/p>\n

\u201cThe bias that is useful to have is contrarianism,\u201d argues Graham. \u201cFootball is too risk averse when it comes to anything. Back in the Spurs days, there wasn\u2019t much scouting of European players \u2014 there was a feeling that foreigners couldn\u2019t do it in the Premier League. That\u2019s mad to think of today.<\/p>\n

\u201cAnd so being a contrarian is to say, whatever the bias is, that you\u2019re willing to go way too far in the other direction, and that will give you an edge. It\u2019s like the Matthew Benham quote of if you\u2019re 1-0 up and down to 10 men in the 90th minute, you\u2019ve got to keep attacking. In some ways, it\u2019s a stupid thing to say, but everyone\u2019s so risk averse that they keep being wrong in the same direction.<\/p>\n

\u201cFind the biases, be biased in the other direction, and even if you\u2019re wrong, because everyone else is wrong in the other direction, that\u2019s going to be an advantage for you.\u201d<\/p>\n

(Illustration: Eamonn Dalton \/ The Athletic; Martin Rickett\/PA Images; Peter Byrne\/PA Wire via Getty Images)<\/p>\n","protected":false},"excerpt":{"rendered":"This is the fourth of a six-part series looking at figures who have played a pivotal role in…\n","protected":false},"author":2,"featured_media":17735,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[563],"tags":[64,63,15420,6257,9663,596,18943,10981,18944,18945,1614,10815,13198,85],"class_list":["post-17734","post","type-post","status-publish","format-standard","has-post-thumbnail","category-football","tag-au","tag-australia","tag-bundesliga","tag-champions-league","tag-championship","tag-football","tag-international-football","tag-la-liga","tag-ligue-1","tag-mens-world-cup","tag-premier-league","tag-serie-a","tag-soccer","tag-sports"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/17734","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/comments?post=17734"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/17734\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media\/17735"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media?parent=17734"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/categories?post=17734"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/tags?post=17734"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}