How an algorithm helped global tennis match-fixing investigation

How an algorithm helped global tennis match-fixing investigation

Buzzfeed’s John Templon has just unveiled 15 months of hard investigative work conducted with the BBC that reveals possible match-fixing at the top echelons of men’s tennis, facilitated by a few handy algorithms.

Using data from the Association of Tennis Professionals along with Grand Slam matches from seven bookmakers, Templon analyzed 26,000 tennis matches from 2009 to 2015 to look for signs that match-fixing might be taking place.

Having worked with sports betting investigators to establish a measure, he was looking for games where the odds of someone losing changed by more than 10 percentage points before it was played.

This identified 39 possible suspects, against which he ran 1 million computer simulations per player to understand just how regularly you would expect them to lose.

From this, he identified 15 players who lost “heavy-betting matches startlingly often,” with one player losing 15 of 16 matches where heavy-betting had taken place against him.

“Betting patterns alone aren’t proof of fixing,” Templon says of his findings. “But it’s extremely unlikely for a player to underperform repeatedly in matches on which people just happen to be betting massive sums against him.

“In fact, according to my simulations, this player would have been expected to lose this many of the matches (or more) less than 1 in 7,500 times, based on bookmakers’ initial odds.”

Buzzfeed says the player identified above is in the world’s top-50 and he is currently competing in the Australian Open.

The investigation is part of a wider look into tennis corruption following a previous one that ended in 2008 and unveiled much the same problems as identified today.

Commenting on the use of algorithms in the newsroom, CEO of data science consultancy Profusion Mike Weston, told TNW: “One of the most intriguing aspects of the tennis fixing story is that a news organisation developed an algorithm to mine betting data to identify suspicious activity.

“But this comes with a note of caution, it is not enough for a journalist to rely on a data scientist or analyst to do the leg work, journalists themselves need to become very data savvy and rigorous.

“Otherwise there is a huge risk of projecting or missing bias in findings, misinterpreting results, or even missing bigger stories. Put simply, it can lead to a journalist making the facts fit the story rather than the other way round.”

Senior tennis officials have denied failing to take action on match-fixing.

How BuzzFeed News Used Betting Data To Investigate Match-Fixing In Tennis [Buzzfeed]

Read next: Pro Linux training is only $19 for a limited time