How data changed gambling

By M.F.

ON JULY 16th, Roger Federer (pictured) triumphed over Marin Cilic to win the men’s Wimbledon tennis championship for the eighth time. It was an expected end to an otherwise unpredictable two weeks, with many top seeds exiting the tournament earlier than anticipated. Despite—or perhaps because of—the unlikely results, punters flocked to the betting windows. Paddy Power Betfair, one of the world’s largest betting groups, saw nearly £1bn ($1.3bn) traded on Wimbledon this year. But it is not just ordinary gamblers who are showing renewed interest in sports betting. In recent years finance and technology types have also been increasingly drawn to the gambling industry: former quantitative traders at investment banks have migrated to the world of sports; job ads asking for machine-learning know-how are not uncommon on bookmakers’ websites. What have complex algorithms got to do with one of the oldest pastimes in the world?

Sports-focused quantitative methods originated in America, where professional managers discovered that they could tease out trends and tactics using data from college-level competitions—a theme explored in “Moneyball”, a 2011 film adapted from a book by Michael Lewis, in which Brad Pitt plays a baseball coach who goes on a record-breaking winning streak after using data-driven analysis to recruit players. Sports like European football, tennis and golf followed. Bookmakers started importing these methods to fine-tune football odds about a decade ago. The job consisted of manually changing win- and goal-expectancy parameters in a spreadsheet as the match progressed. One trader could only oversee one event at a time; resource constraints meant that only the most popular matches were available for live betting.