New – Deep View: An algorithm for better chess commentary

A project by
New – Deep View: An algorithm for better chess commentary

2015: Deep View: An algorithm for better chess commentary

Director’s fellow Maurice Ashley learned how to play chess on the streets of deep Brooklyn. There, lively commentary naturally accompanied intergenerational game play, making what is more globally known as a rather stoic, silence-induced game into something more like a spectator sport. So when Ashley became a chess grandmaster, he felt determined not just to play the game well, but also to commit himself to finding ways of using the stories generated from a game to inspire others.

In November of 2012, MIT Media Lab faculty member Kevin Slavin was attending the Director’s Fellow Program event in New York City when he witnessed Ashley dramatizing a play-by-play, post-tournament debrief with a team of middle school chess players at Hunter College Prep. “Maurice could see into players’ minds in ways that other humans can’t,” said Slavin.“He could change the data on a piece of paper into stories.”

Back at the Media Lab, Slavin assigned student Greg Borenstein to develop an algorithm that could digitize the kind of dramatization that Ashley was doing from his own head. Borenstein, who has taught machine learning at NYU and is also a seasoned illustrator and storyteller, readily took on the challenge. Over the course of several months, Borenstein patched together some code he had written for previous work, working closely with Ashley to understand the inner workings of chess and how the storytelling component of this intricate game could be interpreted by a machine.

Deep View debuted at Ashley’s Millionaire Chess Open event in Las Vegas, Nevada during October of 2014. The grandmasters who played at the tournament used digital chess boards, which sent information about each move they make in real time to Borenstein’s computer. Borenstein, who was sitting alongside the live video production team backstage, then blasted the information he was getting to the switchboard, and the switchboard operators sent the data. Stylized graphics displayed player statistics and strengths like a baseball card, while a ticker below each player’s name showed the percentage chance of them winning.

“Deep View can provide analysis on the fly and can help tell stories,” Slavin says. “It will point out when players deviate from their usual play, as well as their strengths and weaknesses.”