International Chess Day, July 20, looked a little different this year.
This ancient game, with some versions traced back to the seventh century if not earlier,1 underwent a sort of digital renaissance as new players joined the fray to while away the hours during pandemic lockdown. (The Netflix series The Queen’s Gambit didn’t hurt either.)
When The Queen's Gambit aired on Netflix, Chess.com users who studied an online course on the opening sequence of moves named the Queen's Gambit rose by 1,000%. source: Chess.com
Online platform Chess.com, for example, saw tremendous spikes in the volume of play. In March 2019, the site had 1 million active users per day; a year later, the number stood at 5 million.
Many of the world’s tournaments were moved online, replacing face-to-face combat with chess websites and Zoom.
Matt Phelps, an Arbiter (or «chess referee,» he explained) and President of the Boston-area MetroWest Chess Club, said his club’s weeknight gathering had to morph. Its tournaments, typically played with one long game for each player every Tuesday night, switched to multiple «blitz» or fast, casual games played on Chess.com.
«It was the best way to keep in touch with players and keep the club from going into hibernation,» Phelps said.
Good thing digitalization is nothing new for chess. The Internet Chess Club, another site for online play, was founded in 1995 and evolved from BBS bulletin board chess games.2
Today, chess is a data-driven enterprise.
Data Analysis for Chess Improvement
In fact, chess makes a lot of data.
«One thing that drew me to chess is that so much that happens in the game [becomes] data that can be played with after the fact. In other areas, so much is hard to capture,» said David Joerg, a data scientist who works as Chief Learning Officer at Chess.com.
According to Joerg, «Chess is one of the first interactions where everything that matters is available for analysis, both during and after the fact.»
For hundreds of years, chess players have recorded their moves via a written notation system. The modern system is very compact—a sample first move might be recorded as 1.e4—so even a very long chess game, including notes and details such as the time taken on each move, can be stored in a tiny data file of just a few kilobytes. Phelps noted that every major chess game played in a formal tournament is available in a commercial database, «including every position from millions of games.»
And a platform such as Chess.com is recording millions more informal games each day.
Chess.com hosts roughly 10–11 million games every day. source: Chess.com
But what has revolutionized chess is just not these recordings, but the ever improving power of computer algorithms that can analyze chess positions to identify the best available move.
Chess «engines» are now much stronger than even the very best human players.
In fact, Google’s AlphaZero deep-learning chess engine wrote its own evaluation algorithms based solely on examination of a data set of games. It quickly blasted one of the strongest previously developed engines in a match.3 Watching AlphaZero’s play, a prominent English grandmaster named Nigel Short said on social media,4 «I feel like I am in the presence of God.»
While AlphaZero isn’t for sale (and it runs on extremely high-powered hardware), chess players can use other engines to dissect their own play or to understand the games of world-class grandmasters. Computer data analysis is now perhaps the most powerful tool for learning to play chess better.
Some who aspire to improve their chess games have turned their own data into interesting visualizations, looking to spot the opening plans that work best for them or the errors they make most often. Or, sometimes they do it just for fun.5
For better or worse, though, improvement and entertainment are not the only potential purposes for data analysis.
Data Analysis for Cheaters (and for Catching Them)
In July 2019, Latvian grandmaster Igor Rausis was caught using a chess engine for help during a game in the Strasbourg Open tournament in France. Rausis went into a bathroom stall and input the position from his game into a program on his smartphone.
Rausis was eventually stripped of his hard-won grandmaster title and banned from competitive chess.6
How much easier would it be to cheat when playing an online game, sitting at a computer at home, rather than in a live tournament hall?
Much, much easier. So a cat-and-mouse game has developed between those who dare to cheat at online chess and platforms such as Chess.com who have a strong incentive to catch them and shut them out.
Chess.com has a «fair play» team with seven full-time employees devoted to that goal. Early attempts at cheating were easy to spot, as all of a player’s moves would closely match the «best move» recommendations from various chess engines. Now, cheaters try different ways of mixing up their tactics to cover the trail.
Chess.com’s team closed 22,900 accounts in one recent month due to fair-play violations, with 10 of those belonging to internationally titled players. Offenders are banned from opening new accounts, which of course is another area where evasion tactics and the means to find them are constantly evolving.
«Some of the problems we try to solve actually resemble trying to identify spam email,» Joerg said. «Is this behavior from a real person or from automation being applied in a way to try to make it look like a person?»
Though cheating and detection are fascinating exercises in the use of data and analysis, Joerg noted that learning chess and other positive applications still offer a lot more room for creativity. Like any business, Chess.com also uses testing and data to analyze possible new features for its site.
And there is plenty of chess data for anyone to get involved. Joerg himself originally came into contact with Chess.com through a data science contest on Kaggle.
According to Joerg, «Chess.com also has its own API, and I would encourage anyone to hit the API, pull games, and see what they can do with it. I strongly believe we’ve only started to explore what kind of interesting chess questions can be answered.»