More about Cheating in the World Computer Chess Championship

Several people, but most especially Robert Hyatt, the co-programmer of the Cray Blitz computer program, which won the 1986 World Computer Chess Championship, have demanded that I come forward with any evidence of actual cheating at that event.

I have kept my silence about what I saw and observed these past 13 years. I would have been prepared to keep silent about this forever, but since Mr. Hyatt is so insistent that I come out with it, I am constrained to do so.

It was clear to me that there were many instances of cheating at this event. I will start with one example which was not really cheating at all but an honest effort to deal with a difficult problem.

Our second round opponent, Chat, was online with a main frame in I believe Dusseldorf, Germany. Well into the game, our opponent's computer stopped moving. Our opponent called his computer center in Dusseldorf and was told that a big professor at his university had just started running a massive program, a program he ran almost every day, and this tied up system resources, so that all other programs running on the same computer almost completely stopped running.

After a long conversation between my human opponent and the personnel at the computer center (in German of course) it was my understanding that the professor had agreed to hold off running his program. Our opponent started moving again.

However, about twenty minutes later, our opponent again stopped moving. Again, the computer center was called. This time they said the professor was unwilling to wait any longer. Therefore, the programmer decided to change the parameters of the program so that it would run faster. Sure enough, our computer opponent started moving again.

But not for long. Soon the computer stopped and there was another call to Dusseldorf, and then another and another.

I hope you can understand that while I sympathized with our opponent, it was disconcerting to have our opponent changing the parameters of his program, which was actually being operated far away, during the game. We did not protest and we lost the game.

However, the next round opponent of this computer did protest when this same thing happened again. The computer operator was then told that he could not change the parameters in the middle of the game. As a result, when that computer stopped moving, it lost by forfeit.

This has never bothered me because we did lose on the board and I felt that our opponent was being honest and doing the best he could to deal with a problem.

However, on the higher boards, there were other irregularities. In the first place, ours was the only program which displayed an actual chess board on the screen, with the exception of the Hungarian program which turned out not to work and finished last. Because none of the other programs showed an actual chess board, only the computer operators knew for sure what position the computer was thinking about and what moves it actually played.

The surprise of this event was the Rebel program, which scored several upsets and actually could have won the world championship on tiebreaks. In this event, our program, Rex, was one of the few programs to run on a stand alone computer. We were using a Radio Shack TRS model 4. Other programs were running on the Cray, the world's fastest computer, or on other mainframes.

Rebel was one of the few other programs which was running on a stand alone computer. (I believe that David Levy's program was also running on its own computer and Plymate was as well).

Rebel was running on a souped up Apple Computer with a turbo card. The Apple computer top was opened and a fan was blowing on it to stop it from over heating. Several times during the event, the Rebel operator turned off the computer saying that it had overheated.

However, another participant told me that he was once looking over the shoulder of the Rebel operator and saw Rebel play a horrific blunder. The operator did not play that move on the board. Instead, he quickly turned off the computer, saying that it had over heated. When the computer was turned back on, Rebel played a different move.

However, the most serious questions concerned the Cray Blitz program.

Among the 25 programs in the event, Cray Blitz was the only program which was not either there in the tournament hall or online at some remote location. Instead, the arrangement was that Harry Nelson, who I personally knew from the Lawrence Radiation Laboratory in California, was operating a terminal in the playing hall, but he was not online but was essentially calling in the moves to Robert Hyatt, who was in Birmingham Alabama, who in turn was online with a Cray computer in Minneapolis, Minnesota.

Thus, alone among all the computers in the event, and arbiter could not stand in front of the computer terminal and see what moves it was making.

Mr. Hyatt has explained that the reason for this strange setup was that Harry Nelson was not that familiar with the Cray Computer and therefore it was necessary for Mr. Hyatt to interface with it and input the moves.

However, I happen to know for a fact that Harry Nelson was one of the world's leading experts on the Cray Computer. I am sure that Mr. Hyatt is unaware of my connection of Harry Nelson. My connection was that one of the supervisors at the Lawrence Radiation Laboratory where Harry Nelson worked was Nora Smiriga, who was one of my old cohorts from the Berkeley Sexual Freedom League. Nora Smiriga had arranged in mid-1985 for me to go to the Lawrence Livermore Laboratory to see the Cray computers which Harry Nelson was operating. This required a top security clearance, because as you may know the Lawrence Radiation Laboratory was the center for testing nuclear devices, including the neutron bomb, a bomb on which Harry Nelson was working.

Needless to say, in a sensitive project like this one, the United States Government was not going to put in charge a man unfamiliar with how to operate a computer. Nora Smiriga was technically the supervisor of Harry Nelson (which is how she was able to get me in there) but she said that Harry Nelson was one of the "superstars" of the Lawrence Livermore Laboratory.

When I again saw Harry Nelson at the World Computer Chess Championship one year later, to my great surprise, he had forgotten me. (This is one problem I rarely have. Almost nobody forgets me).

One reason why I did not protest at what I saw was out of respect for Nora Smiriga and Harry Nelson.

What I saw and observed was a strange pattern. Frequently, throughout the games, Harry Nelson would lean over to the operator of the opposing computer and say that the operator in Alabama (Mr. Hyatt) wanted to know what move the opposing computer thought was best. Almost all computer chess programs make use of the time when the opponent is thinking by assuming that the opponent will make a certain move and then planning the next move. If the computer guesses correctly what the opponent's move will be, then it will often be able to reply immediately and save time on the clock.

Thus, the opposing computer will always have a move which it thinks best.

What I was able to observe was that Harry Nelson would ask the opponent what the opponent's computer thought was best. He would then relay this information to Mr. Hyatt in Alabama. If the move was the same move which the Cray Blitz thought was best, then that move would be played immediately. If, however, the move was different, then the Cray Blitz would rethink the position and perhaps make the same move or a different move.

You must understand that this small item gave a very big edge. In 1986, many of the programs had bugs. Some bugs would cause the computer to crash. Others would cause the computer to make a horrific blunder. The worst blunder in the 1986 event was made by David Levy's program. It involved a rook which was blocking a pawn from queening. In an otherwise drawn position, Levy's rook moved away from in front of the pawn, allowing the pawn to queen. Mr. Levy did not make a dive for the computer and shut it off, claiming that the computer had overheated. Instead, he made the move on the board and resigned on behalf of his computer a few moves later.

Thus, a simple way of checking to make sure that the next move did not involve a major blunder would give any computer a big advantage.

I finally realized what the Cray Blitz team was doing half way through the final game against HiTech. The next time Harry Nelson asked what HiTech thought was the best move, I tried to approach Hans Berliner and warn him. However, with a wave of the hand, Berliner shooed me away.

I saw Berliner again when he was demonstrating HiTech at the 1986 World Open in Philadelphia. I tried to approach him but again he waived me away.

Finally, at a subsequent event (I cannot remember which one) Berliner came up to me and said that he finally understood what I was talking about, something which he had not understood previously.

Actually, Berliner was complaining about something different, which was that in the final game against HiTech, the Cray Blitz computer had played a much stronger game than it did in the previous rounds, where it played poorly against Bobby and indifferently against Schach 2.7.

If you look at the games and see how improved that the Cray Blitz program was for the final round game, you can see that its playing strength for that last game was at least 200 points stronger than it had been in the previous matches. This certainly suggests the possibility that a strong master had been recruited to help out with that final championship game. Furthermore, Mr. Hyatt has recently mentioned that he used two computers in this event. This raises the question of whether he might have been running two computers simultaneously and using one to check the other.

The question arises of why I have waited these 13 long years to bring this up. The answer is that if I had made these allegations back in 1986 when these events occurred, it would have hurt computer chess and damaged the efforts of the many hard working, dedicated and honest chess computer programmers, including my own partner Don Daily. We had entered this tournament as a learning experience, realizing that we had no hope of winning anything. I certainly did not want to hurt any body or cause and trouble by making these allegations.

In addition, there were national security considerations, because I knew who Harry Nelson really was.

I discussed this with several other participants who had observed similar things to what I had observed. We all agreed that raising these issues would not do anybody any good and would hurt the cause of computer chess and ultimately ourselves.

Now, however, with computers so much stronger than they were then, I see little harm in bringing up these past events. In addition, the Lawrence Radiation Laboratory no longer tests nuclear bombs.

Attached are the five games played by Cray Blitz in this event.

Very Truly Yours,

Sam Sloan

[Event "Computer Wch"]
[Site "Computer Wch"]
[Date "1986.??.??"]
[Round "1"]
[White "Cray Blitz"]
[Black "Awit"]
[Result "1-0"]
[ECO "C42"]
[PlyCount "80"]

1. e4 e5 2. Nf3 Nf6 3. Nxe5 d6 4. Nf3 Nxe4 5. Qe2 Qe7 6. d3 Nf6 7. Bg5 Qxe2+ 8.
Bxe2 Be7 9. Nc3 c6 10. O-O-O Be6 11. Rde1 h6 12. Bf4 O-O 13. Nd4 Bd7 14. Bf3
Re8 15. Ncb5 Na6 16. Nxd6 Bxd6 17. Bxd6 Kh8 18. Kd2 Ng4 19. Nb3 Nxf2 20. Rxe8+
Rxe8 21. Rf1 Ng4 22. Na5 Bc8 23. Nxb7 Nxh2 24. Bxh2 Bxb7 25. Bg1 Nc7 26. Bxa7
Ne6 27. Bh5 c5 28. g3 g6 29. Rxf7 Be4 30. Bg4 Bd5 31. c4 Ng5 32. Rc7 Be6 33.
Be2 Rd8 34. Kc3 Ra8 35. Bxc5 Bh3 36. Bd4+ Kg8 37. Rg7+ Kf8 38. Rxg6 Kf7 39.
Rxh6 Ra5 40. Bf6 Ne6 1-0

[Event "Computer Wch"]
[Site "Computer Wch"]
[Date "1986.??.??"]
[Round "2"]
[White "Bobby"]
[Black "Cray Blitz"]
[Result "1-0"]
[ECO "C29"]
[PlyCount "106"]

1. e4 e5 2. Nc3 Nf6 3. f4 d5 4. fxe5 Nxe4 5. Nf3 Be7 6. d4 O-O 7. Bd3 f5 8.
exf6 Bxf6 9. O-O Nc6 10. Nxe4 dxe4 11. Bxe4 Nxd4 12. c3 Nxf3+ 13. Bxf3 Qxd1 14.
Rxd1 Re8 15. Bf4 c6 16. g4 Bd8 17. Kg2 Be6 18. a4 Rf8 19. Bd6 Rf7 20. Re1 Bd7
21. Rad1 Bh4 22. Re2 Kh8 23. c4 Kg8 24. Rd3 Rd8 25. b4 Rc8 26. Ree3 Bg5 27. Re1
Rd8 28. Bc5 b6 29. Bd6 Bh4 30. Re5 Bf6 31. Re2 Bg5 32. h3 Rc8 33. a5 bxa5 34.
Re5 Bd8 35. bxa5 Bf6 36. Re1 Bd8 37. a6 Bh4 38. Re2 Rd8 39. Bc5 Bg5 40. Rb2 Rc8
41. Rb7 Be6 42. Rd6 Bxc4 43. Rxc6 Rd8 44. Rxf7 Kxf7 45. Bxa7 Rd2+ 46. Kg3 Bd5
47. Rc7+ Kg8 48. Bxd5+ Rxd5 49. Bb6 Rd8 50. Rb7 Rd3+ 51. Kg2 Rd2+ 52. Kf3 Rd3+
53. Ke4 Rxh3 1-0

[Event "?"]
[Site "Computer Wch"]
[Date "1986????"]
[Round "3"]
[White "Cray Blitz"]
[Black "Bebe"]
[Result "1-0"]
[ECO "B70"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 g6 6. Be2 Bg7 7. O-O O-O 
8. Bg5 Nc6 9. Nb3 Be6 10. f4 Nd7 11. Qd3 h6 12. Bh4 Bxb3 13. axb3 a6 14. Bf2 
Nb4 15. Qd2 Qc7 16. Ra4 a5 17. Rd1 Nc5 18. Raa1 Qc6 19. Bf3 Qc7 20. f5 Kh7 
21. fxg6+ fxg6 22. Be3 g5 23. Qe2 Nd7 24. Bg4 Kh8 25. Bf5 Nf6 26. Rf1 Qc6 
27. Rf3 Ng8 28. Qd2 Be5 29. Bxg5 Bg7 30. Rh3 Qc5+ 31. Be3 Rxf5 32. Bxc5 Rxc5 
33. Rf1 Rf8 34. Rxf8 Bxf8 35. g4 Kh7 36. g5 Kg6 37. gxh6 Bxh6 38. Qd1 Nf6 
39. Kg2 Re5 40. Kh1 Rg5 41. Qf1 Kg7 42. Qe2 Rc5 43. Qf2 b6 44. Qg1+ Kh7 
45. Ne2 Nh5 46. Qg4 1-0

[Event "WCCC"]
[Site "Cologne"]
[Date "1986.06.??"]
[Round "4"]
[White "Schach 2.7"]
[Black "Cray Blitz"]
[Result "0-1" adjudicated after Cray Blitz ran out of Cray]

1. d4 Nf6 2. c4 e6 3. Nf3 d5 4. Nc3 Be7 5. Bg5 O-O 6. e3 Ne4 7. Bxe7
Qxe7 8. Nxe4 dxe4 9. Nd2 e5 10. dxe5 Qxe5 11. Qc2 Na6 12. O-O-O Bg4 13. f3
exf3 14. gxf3 Bh5 15. Bd3 f5 16. Rhe1 Rad8 17. Bf1 Qxh2 18. c5 Qe5 19. c6 Kh8
20. cxb7 Nc5 21. Nc4 Qg3 22. Qxf5 Rxd1+ 23. Rxd1 Qxf3 24. Qxf3 Bxf3 25. Rd4
Bxb7 26. Be2 g6 27. Ne5 Re8 28. Ng4 Re4 29. Kd2 h5 30. Nf6 Rxd4+ 31. exd4 Ne4+ 
32. Nxe4 Bxe4 33. Ke3 Bd5 34. Kf4 Bxa2 35. Bd3 Kg7 36. Kg5 Bf7 37. Be4 a5 
38. Bc6 Bg8 39. Be4 Bh7 40. Bd3 h4 41. Bf1 Bg8 42. Bh3 Bd5 43. Bg4 Be4 
44. Kxh4 Kf6 45. Bd7 Bf5 46. Bc6 g5+ 47. Kg3 0-1 {Adjudicated a win for Black}

[Event "Computer Wch"]
[Site "Computer Wch"]
[Date "1986.??.??"]
[Round "5"]
[White "Cray Blitz"]
[Black "Hitech"]
[Result "1-0"]
[ECO "D26"]
[PlyCount "119"]

1. d4 d5 2. c4 dxc4 3. Nf3 Nf6 4. e3 e6 5. Bxc4 c5 6. Qe2 a6 7. dxc5 Bxc5 8.
O-O b5 9. Rd1 Qe7 10. Bd3 e5 11. e4 Nc6 12. Nc3 Bg4 13. Be3 Rd8 14. h3 Bxe3 15.
Qxe3 Bxf3 16. Qxf3 Nd4 17. Qg3 O-O 18. a4 b4 19. Nd5 Nxd5 20. exd5 Rxd5 21.
Bxa6 b3 22. Qe3 Rfd8 23. Bc4 Nc2 24. Qe2 Rc5 25. Rxd8+ Qxd8 26. Rb1 Nd4 27. Qf1
Qd7 28. Ra1 Qc6 29. Bb5 Nxb5 30. axb5 Qb7 31. Ra3 g6 32. Rxb3 Qd5 33. Rb4 Rc2
34. b3 Qd2 35. Rc4 Rb2 36. Re4 Qd5 37. Qc4 Qd1+ 38. Kh2 Rxf2 39. Rxe5 Qd6 40.
Qc8+ Kg7 41. Qc5 Qd2 42. Rg5 Re2 43. Rg4 Qa2 44. Qc3+ Kg8 45. b6 Qa8 46. Qc7
Qf8 47. b7 Re8 48. Rc4 Kg7 49. Rc6 Rb8 50. Qc8 Rxc8 51. bxc8=Q Qb4 52. Qc7 Qxb3
53. Qe5+ Kh6 54. Qf4+ Kg7 55. Qd4+ Kh6 56. Rb6 Qc2 57. Qf4+ Kg7 58. Qf6+ Kh6
59. Qxf7 Qc8 60. Rd6 1-0


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