The world’s richest football league is embracing big data

The world’s richest football league is embracing big data

Aug 17th 2013 |From the print edition

WHEN Liverpool Football Club open the new season against Stoke on August 17th, it is not the 45,000 supporters at Anfield stadium who will be watching most closely. Nor will the band of armchair English Premier League (EPL) fans, spread across 800m homes in 212 countries, have the most detailed view of the game. The EPL has a new breed of observer: sophisticated data firms that meticulously track every movement, using statistics to discover the best way to win.The stakes are high in the EPL. Whopping television audiences make it the world’s richest league, expected to net £3.1 billion ($4.8 billion) in revenues this season. But although EPL clubs make a lot, they spend a lot too. Talent is pricey: players’ salaries absorb 70% of an average club’s takings, and for some that figure is as high as 94%. More than pride is at stake. Since revenues in the league below are a quarter of the EPL’s, relegation can leave clubs with income that falls far short of their outgoings. As several have found out, being relegated can cause a death spiral.

Since the team-sheet can mean the difference between glory and insolvency, clubs employ talent spotters who travel the globe to find the right players at the best price. Until recently, their intelligence gathering was pretty old-fashioned. Brave scouts would spend hours watching matches in the lower leagues in the rain. But now companies such as Opta and Prozone collect reams of helpful data, selling them to the clubs and media for a fee. Pitch-side analysts log every tackle, pass and goal, typically collecting information on 2,000 or so “events” per match. Above the stadium, arrays of cameras track players’ movements, logging their distance, speed and acceleration.

Cold hard numbers offer a fresh view. Take Gareth Bale, last season’s outstanding EPL player. The Welshman’s talents seem easy to spot: he is strong, fast with the ball at his feet, and scores with lots of long distance shots. But crunching the data suggests other attributes that are equally important to Tottenham Hotspur, his club. Mr Bale regularly intercepts opponents’ passes and makes many successful ones himself. The fact that he is good at getting and keeping the ball contributes a lot to his team’s defence. That helps Spurs as much as his goals do.

By scanning lower leagues for players with same attributes, clubs hope to spot cheaper talent (Mr Bale’s mooted sale price is £85m). The ultimate prize is to spot a bargain Bale, a diamond in the rough of lesser leagues. To that end Chelsea, another top club, has data at its fingertips for all players in 15 leagues across the world.

American sports show that this approach can work. In the late 1990s the Oakland Athletics, a threadbare baseball team, were playing badly. But then they started to analyse the huge data sets available in baseball to spot underpriced players, getting them at budget prices. It worked: in 2002 the Athletics enjoyed a record-breaking 20-game winning streak. With reams of new data available, EPL clubs are taking notice. Last year Liverpool recruited a data scientist with a PhD in biological physics.

Still, computers are not about to oust scouts entirely. Sam Green, an analyst at Opta, points out that human observers take into account contextual details that computers don’t. For example, a defender playing for a low-skill team will often hoof the ball upfield at any opportunity. It appears mindless, but this “agricultural” style of play may not imply a lack of talent; his team-mates may simply not be worth passing to. Untangling skill and context is something a computer cannot yet do.