Fantasy football manager

By J.M.F.

The Economist "moneyballs" the English Premier League

IN THIS week’s print edition wediscusshow teams in the English Premier League (EPL) are using “big data” to devise sharper tactics and make shrewder purchases. Specialist sports-data companies such asOptaandProzonerecord each tackle, pass and goal for every player in the EPL, typically totalling some 2,000 “events” per game. Opta has kindly suppliedThe Economist with data for the past two seasons, covering 191 different types of tackle, pass and shot for more than 500 players.

Using topological data-analysis software provided by Ayasdi, a tech start-up, we have visualised the different attributes of players in an experimental interactive chart, below. In a similar way to a Venn diagram, the data is divided into overlapping groups. These groups contain clusters of data—in this case footballers with similar attributes—which are visualised as nodes. Because the groups overlap, footballers can appear in more than one node; when they do, a branch is drawn between the nodes. Some nodes have multiple connections, whereas others have few or none (Wayne Rooney sits alone in our network, for example).

This technique makes it possible to identify what distinguishes football players from their peers without ever watching a match. Take Gareth Bale of Tottenham Hotspur, who was last season's outstanding player in the EPL. Ask a Spurs fan why he is so good and they might cite his ability to score spectacular long-range goals. According to our analysis, it is his not his ability to score from afar that sets him apart from his peers, but the number of successful interceptions and successful passes that he makes in his own third of the field. This suggests that Mr Bale, who has a mooted sale price of £85m ($133m), contributes to the defensive qualities of his team as well as scoring plenty of goals. Using data fromTransfermarkt.co.uk, a fans' website that estimates players' values, we can identify players with similar attributes (ie, those in the same or neighbouring nodes) who could be snapped up for the price of Mr Bale’s left leg. Danny Welbeck of Manchester United and Gastón Ramírez of Southampton appear to represent suitable replacements at less than a quarter of the cost of Mr Bale.

A similar approach was documented in "Moneyball", a 2003 book by Michael Lewis. The general manager of the Oakland Athletics baseball team, Billy Beane, used statistical analysis to find undervalued players according to one or two key metrics. It worked: the Athletics went on a record-breaking 20-game winning streak in 2002. Ayasdi’s software has been applied to sports before, to rethink NBA basketball positions. Baseball is a sport with closed, distinct plays, as well as a history of statistical analysis. Football matches are more dynamic, and the interactions between players are complex. What's more, taking statistics at face value can be misleading: two players may cover the same distance during a season, but one may spend his time running in the wrong direction. Managers are not about to be replaced by mathematicians just yet.

This visualisation is interactive: you can select different areas of the network, search for players and colour the network by players’ positions or attributes such as goals scored (yellow areas indicate more goals per minute, blue areas fewer). We have also highlighted Manchester United's 2012-13 winning squad, in the menu in the top right-hand corner. Try "moneyballing" your own team by selecting a squad with similar attributes to those of Manchester United, at a fraction of the cost.

This visualisation is not optimised for mobile devices, sorry.