NEWS: Shot prediction research paper shortlisted as finalist at 2020 MIT Sloan Sports Analytics Conference

Article by Andy Cooper

Stats Perform, parent company of OptaPro, has announced that a research paper authored by the company’s AI team has been shortlisted as a finalist in the research track at the 2020 MIT Sloan Sports Analytics Conference.

Each year the conference’s Research Paper Competition highlights finalists showcasing cutting-edge research that influences the way the media and professional teams, across various sports, analyse performance.

Co-authored by Dr. Will Gürpınar-Morgan, Dr. Daniel Dinsdale, Dr. Joe Gallagher, Mr. Aditya Cherukumudi and Dr. Patrick Lucey, Stats Perform’s paper — “You Cannot Do That Ben Stokes: Dynamically Predicting Shot Type in Cricket Using a Personalised Deep Neural Network”— introduces a new model for dynamically predicting a batsman’s shot type in One Day International cricket using ball-by-ball Opta event data.

By adopting a personalised deep learning approach, the model takes into account various contextual factors including the match state, a bowler’s delivery trajectory and various personalised metrics for a batsman based on an eight-year archive of international matches. These include measures of a batsman’s ability, aggression and their preferred strike zones when facing different types of bowling.

Taking into account all these factors, the model then assigns probabilities for shot type and their end-location, highlighting the most likely outcome.

This example applies the model to all of Trent Boult’s deliveries from the 2019 Cricket World Cup Final, swapping out each of England’s Top-6 batsman and simulates what types of shot they would play. They are compared to an average batsman to illustrate their shot tendencies.

These insights can inform the strategy and tactics employed by the fielding captain when a specific batsman is on strike, as Stats Perform’s Chief Scientist Dr. Patrick Lucey explains.

“One of the biggest challenges faced in cricket is getting the match-up right between a bowler and a batsman. It can ultimately be the difference between winning and losing.

“This work by our AI team is significant because it can be applied across many different elements of the sport. As well as being valuable to inform pre-game strategies and in-game tactics, it can drive key storylines for broadcasters: highlighting where a player is likely to hit their next shot when taking into account the situation of the game. This can facilitate deeper and more powerful storytelling.”

In the past four years, Stats Perform has reached the finals of the MIT Sloan Best Research Paper Track three other times, winning best paper in 2016 and runner-up in 2017 and 2018.

“It is fantastic to once again see the pioneering work of our AI team recognised by the judging panel at Sloan and I am looking forward to seeing this work presented at the conference later this week,” Lucey adds.

The full paper can be downloaded on the MIT Sloan Best Research Paper Track website here.

Back to analysis