Man vs Machine Redux
World Champion, Vladimir Kramnik , is half way through a match with Deep_Fritz arguably the strongest commercial chess program available. Fritz is running on an 8 processor Compact Computer. This is the program most Grandmaster use for analysis. Hiarcs 8 another excellent program is a more human like opponent and may be equal to Deep Fritz. Both of these programs are stronger than Chessmaster 9000 that beat U. S. Champion Larry Christiansen. Four games into a scheduled eight game match Kramnik leads 3-1 winning twice and drawing the other two games. It appears Kramnik's style is well suited for playing against a computer. He trys to limit the tactical possiblities and focus on a strategic battle. For those interested in more details. Check out the Chessbase site Here is a link to the games if you would like to play through them.
update 10/13/02: The Knight is a popular piece with beginning players the strange move it makes described as one square horizontally or vertically and one square diagonally seems mysterious and quite unpredictable. The Knight and Queen combination is particurlarly powerful. In an equal ending with both players having five pawns and a Queen and a Knight Kramnik simply blundered, apparently overlooking an intermediate check, giving Deep Fritz the win and once again making a match of it. Kramnik continues to lead but the difference is back to one game Kramnik 3 Deep Fritz 2
update 10/15/02: Kramnik with the advantage sacrificed a Knight on move 19, driving Blacks King into the open and great danger. The computer defended magnificintly and Kramnik resigned on move 34. Kramnik said he simplycould not resist a piece sacrifice which could have made this game "the most beautiful of my career" (Kramnik).
The match is now tied Kramnik 3 Deep Fritz 3
update 10/17/02: One game to go and the match is all tied up. Kramnik got the position he wanted out of the opening. Computers don't play as well in closed position's but Kramnik was unable to capitalize on his advantage and in fact allowed the computer to get back into the game. A repetition of moves by Kramnik, and the computer seeing no way to make progress resulted in a draw.
The final game will be played Saturday Kramnik will have the white pieces the question is will he play for a win or try to use the advantage of the first move to secure a draw. A $200,000 dollar decision. He will received a million for victory $800,000 for a draw and $600,000 if he loses.
Match still tied Kramnik 3.5 - Fritz 3.5
update: 10/18/02 : The final game was a boring 21 moves. Looks like to me it may have been agreed drawn in advance. Kramnik not willing to risk $200,000 losing would have cost him. So the match ends drawn 4-4 everyone happy expect perhaps the fans. I found the Christiansen - Chessmaster 9000 match much more enjoyable even though the computer won. Christiansen played Chess Kramnik played a match. Players like Christiansen, Shirov, Ivanchuk, and Morosovic are much more fun to watch than the Kramnik is.
Match drawn Kramnik 4.0 - Fritz 4.0
Here is a link to download the games as a .pgn file or follow this link to play through the games.
note: The first game is listed twice due to a glitch in the viewer, just click on the second copy to load the game. I also suggest turning off animation in preferences. I'll post updates as the results come in, the next game is scheduled to be played tommorrow.
digg
del.icio.us
reddit
Newsvine
FaceBook
Stumble Upon
Comments
Yeah I was wondering when you would comment on this. I also think the way for humans to remain triumphant is to use different boards (3 dimensional stuff, those weird polygon boards) and using the Fischer suggestion of randomly moving around the back pieces...I think that would throw the programmers off for the forseeable future.
Posted by: Philip Shropshire | October 12, 2002 2:41 AM
Does that mean that computers use some sort of pattern matching in determining how to play their opponents?
That's interesting. You could completely throw a computer off very quickly if that was the case...
Posted by: Vinny | October 12, 2002 11:43 AM
Vinny, the computers use both general rules and just brute force calculation. It is the calculation that makes them so formidable. Strong players have been able to exploit positional weakness the programs sometimes create. They are certainly getting better. When I say more human I suppose I'm saying that they are finding the moves human's would most likely play. In the first Kramink - Deep Fritz game the computer retreated its Bishop back to the first rank anticipating a move from Kramnik that would repeat the position and lead to a draw. No human would ever play a move like that , because the human uses more general rules to guide his play. In this case don't move the same piece twice in the opening. Quick development being important to maintaing a balance or gaining an advantage. The computer undeveloped its piece.
Phil I actually had a post ready to go the day the match began, but kept struggling with the Java Viewer trying to get it to work properly, finally a workaround seemed the only solution. Your suggestions of three dimensional chess or Fischer chess are good ones. The computer would still be able to calculate deeper and more accurately than the human, but it would certainly remove the advantage of having an opening library available during play, and the database's of known endings today's programs use to improve their play.
Posted by: Norm | October 12, 2002 4:22 PM
I do not agree with the comment, "Both of these programs are stronger than Chessmaster 9000 that beat U. S. Champion Larry Christiansen."
I am not certain if you are referring to the same Fritz version that recently competed in the ICT3 Leiden Chess Tournament (CSVN), but the King, the engine that Chessmaster 9000 uses, won the tournament, defeating all computer opponents, including Fritz. So, unless this is a radically different version of Fritz, I believe that Chessmaster is actually stronger than Fritz in various conditions. I am not factoring in the multi-processor factors, as these too make a large impact on the performance of any Chess Engine.
Of course, the site is in Dutch; http://www.computerschaak.nl/
Final ranking and round results
Name 1 2 3 4 5 6 7 8 9 P BU SB G1 The King 11w1 9b1 3w1 2b= 8w1 7b1 5w1 4b0 6b1 7.5 47.0 37.25 9 2 Shredder 5w1 13b1 6w1 1w= 3b= 8b1 4w= 7w1 11b1 7.5 45.0 35.25 9 3 Deep Sjeng 4w1 12b1 1b0 5w1 2w= 6b1 10w= 8b1 9w0 6.0 48.5 30.50 9 4 Fritz 3b0 5w= 10b1 6b0 12w1 11w1 2b= 1w1 13b1 6.0 45.0 28.00 9 5 Tao 5.5 2b0 4b= 13w1 3b0 14w1 9w1 1b0 15w1 12w1 5.5 40.0 16.00 9 6 Chess Tiger 8b= 10w1 2b0 4w1 9b= 3w0 bye1 13b1 1w0 5.0 48.0 20.00 8 7 Diep 10b= 14w1 8b= 9w= 11b= 1w0 12w1 2b0 bye1 5.0 42.5 16.50 8 8 The Baron 6w= 11b1 7w= 12b1 1b0 2w0 9b1 3w0 15b1 5.0 42.0 16.50 9 9 IsiChess MMX15w1 1w0 14b1 7b= 6w= 5b0 8w0 bye1 3b1 5.0 40.0 15.25 8 10 SpiderChess 7w= 6b0 4w0 13b= 15w1 bye1 3b= 11w= 14b1 5.0 34.0 13.00 8 11 Ant 1b0 8w0 bye1 15b1 7w= 4b0 14w1 10b= 2w0 4.0 41.0 8.75 8 12 Nullmover bye1 3w0 15b1 8w0 4b0 13w1 7b0 14w1 5b0 4.0 35.5 6.75 8 13 XiniX 14b= 2w0 5b0 10w= bye1 12b0 15b1 6w0 4w0 3.0 36.5 5.00 8 14 31337/Celes 13w= 7b0 9w0 bye1 5b0 15w1 11b0 12b0 10w0 2.5 31.0 2.75 8 15 Praetorian 9b0 bye1 12w0 11w0 10b0 14b0 13w0 5b0 8w0 1.0 31.5 0.25 8
Fritz did defeat Chessmaster during the Tournament, but CM proved itself to be a superior opponent overall in winning the Tournament itself. (Fritz won last year)
Posted by: ChessNut | September 17, 2003 3:21 PM
Thanks for the update. Chessmaster until recently had always been a step behind the others but that has obviously changed, a King indeed. All the programs are incredibly strong as evidenced by the difficulty Grandmasters have in holding their own in competition with the machines.
Posted by: Norm | September 18, 2003 12:10 PM