Topic Rybka Support & Discussion / Rybka Discussion / Rybka vs Ehlvest II - Everything but a Pawn (2788 hits)
Rybka vs Ehlvest II - Everything but a Pawn
On March 6-8, 2007, GM Jaan Ehlvest took on Rybka in an eight-game match organized by IM Larry Kaufman in Potomac, Maryland. In this match, the Estonian grandmaster received pawn odds in every game. It was the first time that a player of Ehlvest's caliber was given such a handicap in a public match, and after a tense struggle, Rybka prevailed by a score of 5.5-2.5.
However, critics have rightly pointed out that in this match, Rybka benefited from three compensating factors which helped to swing the outcome in her favor:
1) Rybka was white in every game.
2) The time control was 45 minutes + 10 second increment, whereas humans benefit from longer time controls.
3) The 8 starting positions included 2 positions in which Rybka was missing one of her rook pawns, which countless computer chess games have shown to be significantly less valuable than non-rook pawns.
In fact, the history of man vs machine play is littered with philosophical arguments about what exactly constitutes equal conditions, which is hardly surprising given the unique asymmetry of such competitions. Given these arguments, and given the lingering questions to which such unequal conditions lead about the exact status of man vs machine, Larry Kaufman decided to find out what would happen if we made to the human player every single significant concession short of giving a material handicap.
The full list of these concessions can be found in Larry's announcement here. Easily the most significant of them is that Rybka's opening book is limited to the first three moves of the game. This simple-looking rule is actually just the tip of a huge philosophical iceberg. It would be easy for us to build an opening book into Rybka, masked (or not) by some output analysis. It would be just as easy for us to tune Rybka in various ways to assimilate modern opening theory. In other words, formalizing this rule is nearly impossible - it only works if the developers are going to follow its spirit. Of course, we will do just that, and as a result Larry's special three-move opening book will need to throw the game into obscure channels in the first three moves at any cost.
The Rybka version which will play in the match is a special private version. You'll probably find that many of her moves will match the choices of the recently-released Rybka 2.3.2a. However, there are a couple of tweaks, including a very nice and quite broad new conception of draw-avoidance which Larry has developed. GM Ehlvest will have to fight hard for every half-point.
On behalf of the entire Rybka team, I'd like to thank:
- Larry Kaufman, himself a Rybka team member, for organizing yet another interesting man vs machine match
- our sponsors, who generously made this match possible
- GM Jaan Ehlvest for taking up the challenge
We wish Jaan good luck and look forward to the games.
Vas
On March 6-8, 2007, GM Jaan Ehlvest took on Rybka in an eight-game match organized by IM Larry Kaufman in Potomac, Maryland. In this match, the Estonian grandmaster received pawn odds in every game. It was the first time that a player of Ehlvest's caliber was given such a handicap in a public match, and after a tense struggle, Rybka prevailed by a score of 5.5-2.5.
However, critics have rightly pointed out that in this match, Rybka benefited from three compensating factors which helped to swing the outcome in her favor:
1) Rybka was white in every game.
2) The time control was 45 minutes + 10 second increment, whereas humans benefit from longer time controls.
3) The 8 starting positions included 2 positions in which Rybka was missing one of her rook pawns, which countless computer chess games have shown to be significantly less valuable than non-rook pawns.
In fact, the history of man vs machine play is littered with philosophical arguments about what exactly constitutes equal conditions, which is hardly surprising given the unique asymmetry of such competitions. Given these arguments, and given the lingering questions to which such unequal conditions lead about the exact status of man vs machine, Larry Kaufman decided to find out what would happen if we made to the human player every single significant concession short of giving a material handicap.
The full list of these concessions can be found in Larry's announcement here. Easily the most significant of them is that Rybka's opening book is limited to the first three moves of the game. This simple-looking rule is actually just the tip of a huge philosophical iceberg. It would be easy for us to build an opening book into Rybka, masked (or not) by some output analysis. It would be just as easy for us to tune Rybka in various ways to assimilate modern opening theory. In other words, formalizing this rule is nearly impossible - it only works if the developers are going to follow its spirit. Of course, we will do just that, and as a result Larry's special three-move opening book will need to throw the game into obscure channels in the first three moves at any cost.
The Rybka version which will play in the match is a special private version. You'll probably find that many of her moves will match the choices of the recently-released Rybka 2.3.2a. However, there are a couple of tweaks, including a very nice and quite broad new conception of draw-avoidance which Larry has developed. GM Ehlvest will have to fight hard for every half-point.
On behalf of the entire Rybka team, I'd like to thank:
- Larry Kaufman, himself a Rybka team member, for organizing yet another interesting man vs machine match
- our sponsors, who generously made this match possible
- GM Jaan Ehlvest for taking up the challenge
We wish Jaan good luck and look forward to the games.
Vas
I asked Ehlvest (before the start of game 1) whether he would prefer his chances under these new handicap conditions or under the previous pawn handicap conditions for this match, assuming the financial terms were identical, and he replied that he thought his chances were better this new way (no pawn handicap). He felt that playing the familiar starting position versus playing the unfamiliar pawn handicap position was the deciding factor. He felt he should at least have reasonable chances in this current match.
Topic Rybka Support & Discussion / Rybka Discussion / Rybka vs Ehlvest II - Everything but a Pawn (2788 hits)
Powered by mwForum 2.22.1 © 1999-2010 Markus Wichitill
