Cheating in the World Computer Chess Championship

I competed in the 1986 World Computer Chess Championship in Cologne, Germany, with my partner, Don Daily, and our program, Rex. Although I was a programmer, Don did almost all of the programming, whereas I put in the chess, deciding that our program would play the Latvian Gambit, for example.

There were several scandals in that event and rampant cheating. I have always wanted to write this story, but have been reluctant to do so, for fear of damaging the reputation of computer chess in general and of hurting the efforts of the many hard working, dedicated and honest computer chess programmers.

 [ Deja News ] Search Discussion Groups For:

Now, however, with these events far in the past of 13 years ago, plus the strongest programs are now stronger than all but the top grandmasters, I feel that it is time to tell the story of what really happened at that 1986 event.

In order to do this, I need the games of the top two programs - Cray Blitz and Hi Tech. I cannot find them online anywhere. Can anybody e-mail them to me?

Let me just summarize by saying that Cray Blitz did not deserve to win that tournament. The winner should have been the Hi Tech program of Hans Berliner.

My own program finished near the bottom in 21nd place. However, although it did not really matter since we would have been near the bottom anyway, we too were the victims of cheating. In fact, I found this so upsetting that I dropped out of computer chess altogether. I never chess programmed again. My partner, Don Daily, however, went on to become one of the top programmers in the field.

One of the games where we were cheated was against Awit, a Canadian program. Our program, Rex, was almost the only program in the tournament which had an actual chess board on display. Thus, anybody could look at our computer and see the actual position. Most other programs reported moves in machine gibberish, which only the programmer understood for sure, and this opened the door to much of the cheating.

Our game against AWIT pitted two of the weakest programs in the tournament against each other. In this game, both programs brought out queens early, grabbed pawns, and neglected development. By move 16, the game appeared to be headed for a draw, because both sides had kings exposed to queen checks.

International Master Michael Valvo and Grandmaster Hort were the MCs for the event which was held in the middle of a computer trade show. Valvo, announcing, came over to our demonstration board and, looking at the position, said to me privately, "Both of you guys are crazy."

Then, to the public, Valvo, said "This game seems headed for a draw. There is going to be an endless series of queen checks. There is no way for either side to escape."

Then, Valvo paused for a few moments and said, "There is only one move to keep this game going, and that is for white to play Na3."

Suddenly, only about 30 seconds after Valvo uttered these words, my opponent's computer, which was online at Calgary, Alberta, Canada, crashed! My human opponent stopped the clock and said that it would take a half hour before his computer in Canada could start up again. This was allowed by the arbiter, Valvo. Then, my human opponent, Tony Marsland, proceeded to set up the board by hand and finally started the clock again. Amazingly, his program played 17. Na3 almost instantly!

His program, AWIT, was a selective search program, which meant that his program did not analyze every possible move but only selected the top six candidate moves and analyzed those. With a variety of queen checks available, it is almost guaranteed that 17. Na3 would not be one of the top six candidate moves, especially since his program had been neglecting development up until that time.

Later in the game, Marsland asked to stop and start his computer again. This time, he said that when he set up the computer before it played 17. Na3, he had forgotten to tell it that it could still legally castle, and as a result his computer was not castling. However, Valvo, without thinking of the implications of this, casually told him, "What difference does it make. Your computer is winning anyway."

As it turned out, my program finished 21nd and his program finished 20th. Had this game been a draw, as it otherwise probably would have been, our positions might have been reversed. As I said before, this was of no great moment. However, what was of great moment was how the 1986 World Computer Chess Championship was decided by cheating on the top boards, which I will explain when I get the games.

Here is our game against AWIT.

Sam Sloan

[Event "World Computer Chess Championship"]
[Site "Koln, Germany"]
[Date "1986.??.??"]
[Round "4"]
[White "Awit "]
[Black "Rex"]
[Result "1-0"]
[ECO "A01"]

1. b3 e5 2. Bb2 Nc6 3. e3 d5 4. Bb5 Qg5 5. Nf3
Qxg2 6. Rg1 Qh3 7. Bxe5 Bg4 8. Rg3 Bxf3 9. Qxf3
Qh6 10. Qxd5 Nge7 11. Bxc6+ bxc6 12. Qf3 Nd5 13. Qg4
Qxh2 14. Bxg7 Bxg7 15. Qxg7 Ke7 16. Qe5+ Kd7 17. Na3
Ne7 18. Qd4+ Ke6 19. Qe4+ Kd7 20. Rg7 Qh5 21. Nc4
Ke8 22. Ne5 f5 23. Qg2 Qh6 24. Nxc6 Nxc6 25. Rxc7
Kf8 26. Rxc6 Qg7 27. Rf6+ Qxf6 28. Qxa8+ Kf7 29. Qxa7+
Kf8 30. d4 Ke8 31. Ke2 Kf8 32. c4 h6 33. a4
Rg8 34. a5 Rg7 35. Qc5+ Re7 36. a6 f4 37. a7
f3+ 38. Kf1 Kg8 39. a8=Q+ Kh7 40. Qh5 Rf7 41. Ra6
Kg7 42. Rxf6 Rxf6 43. Qe5 h5 44. Qxf3 1-0
-------------------------------------------------------------


Robert Hyatt, the programmer for Cray Blitz in that event, responds:

In rec.games.chess.computer Sam Sloan wrote:
Cheating in the World Computer Chess Championship

I competed in the 1986 World Computer Chess Championship in Cologne, Germany, with my partner, Don Daily, and our program, Rex. Although I was a programmer, Don did almost all of the programming, whereas I put in the chess, deciding that our program would play the Latvian Gambit, for example.

There were several scandals in that event and rampant cheating. I have always wanted to write this story, but have been reluctant to do so, for fear of damaging the reputation of computer chess in general and of hurting the efforts of the many hard working, dedicated and honest computer chess programmers.

Now, however, with these events far in the past of 13 years ago, plus the strongest programs are now stronger than all but the top grandmasters, I feel that it is time to tell the story of what really happened at that 1986 event.

In order to do this, I need the games of the top two programs - Cray Blitz and Hi Tech. I cannot find them online anywhere. Can anybody e-mail them to me?

Let me just summarize by saying that Cray Blitz did not deserve to win that tournament. The winner should have been the Hi Tech program of Hans Berliner.

My own program finished near the bottom in 22nd place. However, although it did not really matter since we would have been near the bottom anyway, we too were the victims of cheating. In fact, I found this so upsetting that I dropped out of computer chess altogether. I never chess programmed again. My partner, Don Daily, however, went on to become one of the top programmers in the field.

One of the games where we were cheated was against Awit, a Canadian program. Our program, Rex, was almost the only program in the tournament which had an actual chess board on display. Thus, anybody could look at our computer and see the actual position. Most other programs reported moves in machine gibberish, which only the programmer understood for sure, and this opened the door to much of the cheating.

I have never seen such total rubbish.

Our game against AWIT pitted two of the weakest programs in the tournament against each other. In this game, both programs brought out queens early, grabbed pawns, and neglected development. By move 16, the game appeared to be headed for a draw, because both sides had kings exposed to queen checks.

International Master Michael Valvo and Grandmaster Hort were the MCs for the event which was held in the middle of a computer trade show. Valvo, announcing, came over to our demonstration board and, looking at the position, said to me privately, "Both of you guys are crazy."

Then, to the public, Valvo, said "This game seems headed for a draw. There is going to be an endless series of queen checks. There is no way for either side to escape."

Then, Valvo paused for a few moments and said, "There is only one move to keep this game going, and that is for white to play Na3."

Suddenly, only about 30 seconds after Valvo uttered these words, my opponent's computer, which was online at Calgary, Alberta, Canada, crashed! My human opponent stopped the clock and said that it would take a half hour before his computer in Canada could start up again. This was allowed by the arbiter, Valvo. Then, my human opponent, Tony Marsland, proceeded to set up the board by hand and finally started the clock again. Amazingly, his program played 17. Na3 almost instantly!

And this would suggest cheating? First, stopping the clock was the correct decision. Second, awit could play almost anything, it was _highly selective_ and could play good or bad moves almost on a whim, as any selective program does...

You said you were going to show how you were 'cheated'. You have only shown a lack of class and a lot of 'sour grapes'. You could have easily protested _after the game_ and had Mike have Tony restart Awit from the questioned position to see if it played that move on its own.

His program, AWIT, was a selective search program, which meant that his program did not analyze every possible move but only selected the top six candidate moves and analyzed those. With a variety of queen checks available, it is almost guaranteed that 17. Na3 would not be one of the top six candidate moves, especially since his program had been neglecting development up until that time.

How could you 'almost guarantee' anything? That's baloney, knowing how Awit played (It was around for well over 10 years as either Wita or Awit.)

Later in the game, Marsland had to stop and start his computer again. This time, he said that when he set up the computer before it played 17. Na3, he had forgotten to tell it that it could still legally castle, and as a result his computer was not castling. Therefore, with the permission of Valvo and over my objection, since I could easily see what Marsland was doing, he restarted his computer and set up the board a second time.

What would be the basis of your protest. Errors are supposed to be corrected when they are found. You might have requested that the game be backed up to the Na3 move again and restarted from there with the castling status set up correctly, but none of this suggests 'cheating'.

As it turned out, my program finished 22nd and his program finished 21st. Had this game been a draw, as it otherwise probably would have been, our positions might have been reversed. As I said before, this was of no great moment. However, what was of great moment was how the 1986 World Computer Chess Championship was decided by cheating on the top boards, which I will explain when I get the games.

You are making baseless statements. Hans Berliner protested the Cray Blitz vs HiTech game claiming that in one position _no_ computer would play the move we played. We put a bishop on a square where a knight could take it, but if so, we would play pxn and create a powerful passed pawn. His program took the bishop, we took the knight and won the game with that pawn. Later Ken Thompson noted that Belle also played that same move. At the ACM event that year, David Levy wanted to go over the game. Harry Nelson contacted the Cray computer center to set up time (we couldn't play in the ACM event because it was too close to the WCCC and we couldn't get machine time for both.) He had them (with Levy involved) restore the exact copy of Cray Blitz executable that was on the machine when the backup was done that evening right after our game with HiTech ended. They then went over the game, and Cray Blitz played _every_ questioned move, matching the log file I had already sent to David the night of the game after the protest was made.

He found _nothing_ wrong in our game, no move that had a different score, no move where the program wouldn't play the move played in the game. In short, the protest was dismissed as baseless.

I assume you have _other_ data. But since you need game scores to show that cheating occurred, I smell sour grapes and baseless accusations once again.

Here is our game against AWIT.

Sam Sloan

[Event "World Computer Chess Championship"]
[Site "Kohn, Germany"]
[Date "1986.??.??"]
[Round "4"]
[White "Awit "]
[Black "Rex"]
[Result "1-0"]
[ECO "A01"]

1. b3 e5 2. Bb2 Nc6 3. e3 d5 4. Bb5 Qg5 5. Nf3 Qxg2 6. Rg1 Qh3 7. Bxe5 Bg4 
8. Rg3 Bxf3 9. Qxf3 Qh6 10. Qxd5 Nge7 11. Bxc6+ bxc6 12. Qf3 Nd5 13. Qg4 Qxh2 
14. Bxg7 Bxg7 15. Qxg7 Ke7 16. Qe5+ Kd7 17. Na3 Ne7 18. Qd4+ Ke6 
19. Qe4+ Kd7 20. Rg7 Qh5 21. Nc4 Ke8 22. Ne5 f5 23. Qg2 Qh6 24. Nxc6 Nxc6 
25. Rxc7 Kf8 26. Rxc6 Qg7 27. Rf6+ Qxf6 28. Qxa8+ Kf7 29. Qxa7+ Kf8 
30. d4 Ke8 31. Ke2 Kf8 32. c4 h6 33. a4 Rg8 34. a5 Rg7 35. Qc5+ Re7 
36. a6 f4 37. a7 f3+ 38. Kf1 Kg8 39. a8=Q+ Kh7 40. Qh5 Rf7 
41. Ra6 Kg7 42. Rxf6 Rxf6 43. Qe5 h5 44. Qxf3 1-0

Robert Hyatt Computer and Information Sciences
hyatt@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170

Here are links:

Sam Sloan's Chess Page

Return to my Home Page


Contact address - please send e-mail to the following address: Sloan@ishipress.com