Fritz and Shredder win Computer Chess Championship

Once again all four professional programs won their games, quite easily and without much ado. In round ten Fritz cut Chinito to ribbons, but in the final round Fritz seemed to have a dead draw against Parsos. Most people had already began to celebrate Shredder (which was obviously winning) as the new world champion, when the new positional algorithms of Fritz started to do their magic and the program slowly ground down its German opponent.

Rudolf Huber of Parsos (left) does not understand what the positional magic of Fritz is doing to his drawn position

Deep Junior beat Diep in the penultimate even though the Dutch program for a time had two extra pawns. Brutus easily disposed of Nexus but had to do actual work to beat Quark.

With GM Peter Wells playing a league game (against a carbon-based opponent) it was left to ICGA president David Levy to comment on the games

In round ten Shredder deftly outplayed Parsos, investing two pawns and then a piece for an irresistible kingside attack. In the last round the many-time world champion had white against the trailing amateur program Jonny and duly went into the attack.

Shredder - Jonny [B80] WCCC 2003 Graz (11), 29.11.2003
1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 d6 6.Bg5 Be7 7.f4 0-0 8.Qf3 e5 9.Nf5 Bxf5 10.exf5 e4 11.Qh3 h6 12.Bh4 Qc7 13.g4 d5 14.g5 hxg5 15.fxg5 Nh5 16.0-0-0 Rc8 17.Bg3 Nf4 18.Bxf4 Qxf4+ 19.Kb1 Bxg5 20.f6 Nc6 21.fxg7 Kxg7 22.Be2 d4 23.Rhf1 Qh4 24.Qf5 dxc3 25.Rg1 Kf8 26.Rxg5 Ne7 27.Qf6 Qh7 28.Rd7 Ng6 29.Rxb7 Qg7 30.Qd6+ Kg8 31.Rxf7 Qxf7 32.Rxg6+ Kh7 33.Rg4 Rab8.

At this stage Shredder was displaying an advantage of over ten pawns, it was on the verge of announcing mate. The game continued 34.Rh4+ Kg8 35.Rg4+ Kh7 36.Rh4+ Kg8 37.Rg4+ Kh7. At this point somebody said that Jonny had claimed a three-fold repetition, but everyone assumed that was a misunderstanding. However, spectators on the Playchess.com server soon ascertained that the notation posted there indeed did contain a position that had occurred three times, with the same side to move. You can see for yourself by loading the game on our Javascript board and clicking on the moves 33...Rab8, 35...Kh7 and 37...Kh7. The board position doesn't change.

In the meantime the game had continued, with Shredder deviating with 38.Bc4 Rxb2+ 39.Ka1 Rxc4 40.Rh4+ Kg8 41.Qd8+ Qf8 42.Rg4+ Kf7 43.Qd7+ Qe7 and Stefan Meyer-Kahlen's program was already announcing mate. This duly came with 44.Rf4+ Kg6 45.Qxe7 Rxa2+ 46.Kxa2 Ra4+ 47.Kb3 Rb4+ 48.Kxb4 a5+ 49.Kxc3 a4 50.Qf6+ Kh5 51.Rh4# 1-0.

Stefan Meyer-Kahlen (left) discusses Shredder's game with Johannes Zwanzger, the author of Jonny.

But what about the three-fold repetition? After making sure there was no error in the notation the ICGA jury retired to discuss the matter. Apparently the Shredder interface contained a bug which allowed it to repeat postitions in a totally winning position. The program Jonny had seen this and claimed the draw (while displaying a 0.00 score). But as Johannes Zwanzger later said on the Playchess.com server: "I did not want to draw the game in this way against Stefan, just because his program has a bug." So he had simply executed the move on the board, entered Shredder's reply and continued with the game. [Here's an eye-witness account of what transpired]

This is a very sporting gesture by the rookie programmer, one that brought him the approving applause of many spectators. But his action does create some problems for the ICGA. The first rule in computer chess has always been that the human operator must be completely passive and may not interfere in any way with the outcome of the game. Certainly he may not simply hand over half a point to the opponent, however much he may like him or admire the performance of his program. In this case his action in fact pushed Shredder into the joint lead with Fritz, making a playoff on Sunday necessary to decide the world champion.

Press conference with main arbiter Prof Jaap van den Herik (right)

After debating the matter for a considerable period of time Prof Jaap van den Herik appeared and read the decision of the committee given below. You could sense that the world computer chess body was uncomfortable with its decision. His speach and the Q&A session could be heard by visitors to the Playchess.com server.

In its decision the ICGA confirms that the Jonny program had announced its move and stated on the screen “info” and “dreifache Stellungswiederholung” (“information” and “threefold repetition of position”). But, said Jaap van den Herik, this is different from "announcing its intention of making the move and displaying wording to the effect that it was claiming a draw," as the FIDE rules would requrie. The "Info" display only meant that the program was supplying status information, not claiming a draw.

Of course computers have always announced threefold repetitions in a Windows alert box, we know of none that goes to the arbiter (or instructs its operator to do this in its stead) to formally claim the draw. Were all previous claims by computer programs illegitimate?

The ICGA also ruled that because the move had been completed on the board it was no longer possible for the operator to claim a draw. But that would imply that other errors committed by operators on the physical board would have similar irrevocable consequences. What if an operator touches a piece or executes a move that was not displayed by his program? Do you force the program to play a move with the piece that was touched or accept the move the operator erroneously executed on the board?

The positive side is that the ICGA has introduced an endearing human element into computer chess, something the organisation had for many decades studiously tried to eliminate from their tournaments. Minor program bugs will no longer automatically lead to losses or disadvantages in the game, it is up to the operator to decide how to act on such matters.

Here are all the results from the Computer Chess World Championship 2003 in Graz:

Final Standings

• All games so far in ChessBase format (with computer protocols)

• There will be a playoff between the two leading programs Fritz and Shredder on Sunday, November 30 at 9 a.m. CET.

All games are being transmitted live on the Playchess.com server. This includes audio commentary by GM Peter Wells and video impressions from the tournament hall.

In order to follow the games you can use Fritz or any Fritz-compatible program (Shredder, Junior, Tiger, Hiarcs) to follow the action, or download a free trial client here.

Links

• Official web site
• Schedule
• Video trailer (WMV format) by Imagewerk

ICGA
INTERNATIONAL COMPUTER GAMES ASSOCIATION

9, Springfield Avenue,
London N10 3SU,
England.

Telephone: +44 (0) 208 883 5180
e-mail: dlevy@msoworld.com

David Levy - President

Graz, Austria.
November 29th 2003

The game Shredder vs Jonny, 11th round, World Computer Chess Championship

During the game Shredder vs Jonny in the final round of the World Computer Chess Championship a position arose for the third time in the game.

Rule 5 of the World Championship is: “Unless otherwise specified, rules of play are identical to those of human tournament play. If a point is in question the Tournament Director has the right to make the final decision.” There is no mention in the World Championship rules of the procedure for claiming a draw by threefold repetition and so the rules of tournament chess apply.

Under the FIDE rules, in order to claim a draw by threefold repetition a player must announce his intention of making the move that leads to a threefold occurrence of the same position and must make the claim, but without actually making the move on the board. If he makes the move on the board this nullifies his claim.

When the relevant position arose in the Shredder vs Jonny game the Jonny program announced its move and stated on the screen “info” and “dreifache Stellungswiederholung”, meaning “information” and “threefold repetition of position” respectively. The operator of the Jonny program then made the program’s move on the board, pressed the clock and went to call the Tournament Director, Jaap van den Herik. When Professor van den Herik arrived at the board he could see that the move leading to the repetition had been completed on the board.

Some discussion ensued. Profesor van den Herik ruled that because the move had been completed on the board it was no longer possible for the operator to claim a draw.

This game became the subject of much discussion on the Chessbase site and will doubtless be discussed at length elsewhere. The ICGA would therefore like to make it absolutely clear why Professor van den Herik’s ruling was correct.

A draw could have resulted in one of two ways. Firstly, the program could have claimed a draw using the correct procedure, announcing its intention of making the move and displaying wording to the effect that it was claiming a draw. (This is not the same as the statements “information” and “threefold repetition of position”.) In this case the operator would have been obliged to advise the Tournament Director that his program had claimed a draw. The Tournament Director would then have verified the claim by examining the move record of the game.

The second way is that, after the program’s statement, the operator could have gone to the Tournament Director to claim a draw, without making the move on the board and pressing the clock. The point here is that the move is not complete until it has been made on the board, the hand has quitted the piece and the clock has been pressed.

What in fact happened is that neither the program nor the operator made a valid claim for a draw. Under these circumstances the fact that the game was continued by the operator completing the move on the board nullifies the subsequent claim for a draw. The Tournament Director’s decision was therefore correct.

It is most unfortunate that this incident occurred in a game by one of the programs that tied for first place in the tournament. Those who have not seen the game may be interested to know that Shredder was in a completely winning position when, for an as yet unknown reason, it allowed the repetition. When the game continued Shredder quickly won.

David Levy
[President]

An eye-witness description of what happened

By Deep Junior programmer Amir Ban

In the final phase of the game Shredder, in an easily won position, became indecisive due to bugs, and finally stepped into a three-fold repetition while showing a mate score. On Jonny's screen with the Chessbase interface the threefold repetition message popped-up, and the game was marked as a draw.

Jonny's operator realized that Shredder had thrown away the game, and the championship, and out of chivalry did not want to accept that. He went to the TD v.d. Herik and asked for permission to continue playing.

However, the TD did not hear or understand the request, and told him to wait until he comes by the board. When he arrived, Jonny had already played the repetition move, and Shredder was pondering. The TD, still thinking that Jonny was trying to claim a draw, ruled that as a move was played the draw cannot be claimed. While the game continued this was discussed by the viewers, and brought again to the attention of the TD, who said he will consider the matter when the game is over.

When the game was over, the TD with other ICGA officials questioned the Jonny and Shredder programmers about what had happened, inspected Jonny's chessbase log, and talked to spectator programmers, including myself. Then they ruled the Shredder's win stands, and called a programmer's meeting to announce and explain the decision.

At this stage they still were not aware that Jonny's operator wanted to continue rather than claim a draw. However, during the discussion Jonny's operator came on stage and told frankly that his question to the TD when the 3-fold repetition pop-up occurred was whether he is allowed to ignore it and continue.

I said at the meeting that in this case the ruling is not valid, because it is not the case that Jonny erred in claiming a draw, but the opposite: the operator did not want to claim it, and this is something he should not be allowed to do.

Suppose the TD had understood his question: "I can claim a draw now, but I request permission to go ahead and get mated". The obvious answer by the TD is: "No way. You are not allowed to lose on purpose".

The TD, perplexed by this new twist, said that while possibly Jonny's operator may be censured, his intentions do not change the technical chain of events, so the ruling stays.

My opinion: In a human game refusing to claim a draw out of chivalry is something that is within the rules. However, in a computer game the operator should not be allowed to make decisions that are against the interests of the program. An equivalent would be an endgame KNP vs. K, where the stronger side, due to a bug, loses the pawn, but the opponent, rather than taking the pawn, chooses to resign. No ICCA TD would allow such a resignation.

Posten in the Computer Chess Club