Cheating In Chess

Colonel Stok: Do you play chess?
Harry Palmer: Yes, but I prefer a game with a better chance of cheating.—Funeral in Berlin

Ken Regan is well known to us all, and is the co-author of this blog. He is in the Department of Computer Science and Engineering, University at Buffalo (SUNY)–as part of the theory group. He is also an International Master from the World Chess Federation (FIDE).

Ken’s Trip

Ken has used his ability in theory with his expertise in chess to study how to detect cheating in chess. Unfortunately people do currently cheat in chess, and so detecting them is an important problem for FIDE. Ken has just spent some time in Bologna, Italy at the International Chess Federation Fair Play Commission’s meeting. The group’s goal is: There is an increasing demand for fair play experts during chess events, and we want to provide the organizers with professionals who know how to collect evidence, apply law, use the state-of-the-art detection tools. And do it in the way that players are protected and public perception of how important the fair play is improved.

Ken is on the way back from Italy and soon will explain in detail what is the latest about cheating in chess.

Cheating In The Past

In the past cheating at chess started with machines like the famous Turk of 1770. This was a machine that claimed to play chess without human help. Even just playing legal moves would have been impressive, but the Turk played strong chess.

The secret to the Turk is it used a hidden human chess player to make its moves. The player was hidden inside the machine. The audience was allowed to examine the machine by opening and looking at parts of the machine. The audience member could only examine one part of the Turk at a time. The trick was that the player could move from one part of the Turk to another. At each time some parts of the Turk would be visible, but some part stayed invisible. This meant the player could move from one part to another of the Turk. This fooled the audience and made it seem “magical” that the mechanical system could play chess.

Cheating Now and in the Future

Cheating at chess became interesting again when computer programs started to play strong chess—cheating. The fundamental insight was that once programs could play master level chess, players could cheat and play at this level. The idea was simple: A player would not look at the board and select the next move. They would instead ask the program what move should they do. They would then use that move.

Of course a serious issue was how could the player ask a program to make the next move? If the game they were playing was offline, then the player should be able to use their laptop to remotely run the program. If the game was played in some public place, then the player might have to work harder to run the program. But a player could perhaps still do this without being detected.

Technology has been used by chess cheaters in several ways. The most common way is to use a chess program while playing chess remotely, such as on the Internet or in correspondence chess. Rather than play the game directly, the cheater simply inputs the moves so far into the program and follows its suggestions, essentially letting the program play for them. Electronic communication with an accomplice during face-to-face competitive chess is a similar type of cheating; the accomplice can either be using a computer program or else simply be a much better player than their associate.

Open Problems

The main open problem is still: How well can cheating be detected? The basic idea is simple: Suppose that the player named Carol plays a game in some tournament. We wish to determine whether Carol played her own moves or whether she used moves of some program. How can we do this? There are several issues that make this hard.

1. Carol may stay on “book” at the beginning. How do we factor this in?
2. Carol may have used a program that is secret. Does this effect the detection?
3. Carol may follow a program for all moves or for some of the moves. Does this make the detection more difficult?
4. Carol may already be a strong player. Suppose she uses the program to play at a stronger level. Is this even more difficult to detect?

Ken will update us on the latest views of these and other issues.