Rybka v GM Perelshteyn 1.5-0.5

[Graham Banks]

Two game match between GM Eugene Perelshteyn (FIDE 2555, age 28) and Rybka (slightly modified from Rybka 3) with a triple handicap as follows:

1. GM gets White in both games.
2. Ten to one time handicap ratio! GM plays the standard FIDE 90'+30" increment, Rybka plays 9'+3" increment.
3. No thinking on opponent's time by Rybka ("Ponder" turned off).

Hardware: Octal computer at about 3 GHz.
The time handicap (with ponder off) should roughly simulate playing with equal time (but with ponder off) on a sp 32 bit machine at around 2.4 GHz.

The result: 1.5-0.5 to Rybka.

The games can be played through here:
http://www.chessok.com/broadcast/live.html

[Graham Banks]

Sorry - originally posted incorrect result. Now corrected.

I'm sure I'll be accused of being a Rybka fanboy for posting this.

[Marc MP]

Not for mistakenly reporting the score Graham, but I don't agree about the time control being an handicap. Had we been in 1995, a good hardware was more than 10 time slower than an octal computer today. I don't think the ponder off will offset that. The only real disadvantage I see is that Rybka is playing with the black pieces for both games. What I find surprising is that a 3200 elo program gives a draw against a mere 2550 player.

[Dirt]

Two game match between GM Eugene Perelshteyn (FIDE 2555, age 28) and Rybka (slightly modified from Rybka 3) with a triple handicap as follows:

1. GM gets White in both games.
2. Ten to one time handicap ratio! GM plays the standard FIDE 90'+30" increment, Rybka plays 9'+3" increment.
3. No thinking on opponent's time by Rybka ("Ponder" turned off).

Hardware: Octal computer at about 3 GHz.
The time handicap (with ponder off) should roughly simulate playing with equal time (but with ponder off) on a sp 32 bit machine at around 2.4 GHz.

The result: 1.5-0.5 to Rybka.

The games can be played through here:
http://www.chessok.com/broadcast/live.html

A nice fortress in the second game. "Father" should be pleased.

[swami]

Rybka plays worser with pawn handicap than it would with the (present)time handicap. So I'd propose that Rybka should get 5+2 time for the whole game instead.

[shiv]

The second game is not a deliberate fortress attempt by the GM. The position was equal and the draw logical. The game was long since the computer did not recognize the draw at an earlier stage.

When strong humans play solidly with the white pieces it is hard to win for black.

[Marc MP]

Rybka plays worser with pawn handicap than it would with the (present)time handicap. So I'd propose that Rybka should get 5+2 time for the whole game instead.

Hi Swami,

I agree that being a pawn down with the same hardware (with no time disadvantage) is a greater disadvantage.

But I think that presenting the match as a "Time handicap match" (as it is at: http://www.chessok.com/broadcast/live.html ) for Rybka is a bit of a mismonner. My pentium 3.2Ghz prescott is roughly 15 times slower than an 8-way intel accoring to the Fritz benchmark (http://www.jens.tauchclub-krems.at/dive ... marks.html).

Had Rybka played on my hardware instead, to keep the computing power approximately the same, it would have needed to have a time advantage (i.e something 105min + 105sec) for the game! Then the GM would give Rybka a time advantage!

Well marketing issues I guess.

Nonetheless I think that limiting the computer power for the engine is the way to go to keep these human-engine match interesting. My personal preferences for the upcoming match with Milov would have been: (with equal time and material, and white and black pieces the same) Pentium 4 2.0Ghz (the old standard from CEGT). If we can keep the computing power constant, we will have a better idea of the progress of the pure engine strenght. As it is now, it mixes engine improvements with hardware improvement.

[Karmazen & Oliver]

THE FIRST party-game ? (for rybka obviously ), is quite bad... faulty.

only 3 seconds ?¡ the alone human player has needed to tolerate the rybka lunges the first ones 20 or 30 plays-moves, later, with similar control of times, a GM never, and I say it well, it can never lose a game,

incorrect strategy, that control of time prevents to rybka to think in more of 15 plys~18... possibly open game fatale... the book of gm is ... ok. no more.