Author: Odd Gunnar Malin
Date: 05:52:39 02/10/03
Go up one level in this thread
On February 10, 2003 at 07:49:42, Maurizio De Leo wrote: >On February 10, 2003 at 05:48:42, Odd Gunnar Malin wrote: > >>If you did read this before the match, what were you thinking? >>Ok, they don’t thrust the program when it comes recognise a drawn position so >>they want a human/chess expert to take this decision (do this 'move') on the >>behave of the program. Isn't obvious then that this team should only see the >>chessboard under the game and not any analyses or even the score output from >the playing program. > >I read this before the match, because it was published on the chessbase news and >also advertised here in CCC. > >When I read it I tought it as a "kindness" to the human player. >It is clear that the draw agreement can be coded inside the program : for >example "accept only if down 0.x pawn or more for more than y moves". This >feature is implemented in every commercial program, also Junior 7. >But we have seen in game 4 that computers can be dumb in understanding dead >draw, and can play them on for dozeens of moves, thinking to be ahead. >From this the idea of delegating the decision to the human. You only points on the error I was referering to, and why they didn't thrust the engine. If you have to play 50 moves with black in an endgame with f.ex. wrong bishop color. (white pawn on h3, black pawn on h4 and h5, white have a 'white' bishop and you holding h8 with your king). Isn't the next stop for the engine the recycle bin. > >But there was a problem. Kasparov is 2800 and Deep Junior is supposed to be a >worthy opponent : how could a 1800 elo human operator decide about possible >outcome of a super-Gm game ? > >So the "team" (including a Gm) decides, checking of course what the real player >thinks (this means "what's junior score"). >It is clearly stated in the rules that they couldn't offer draw if Junior >evaluation wasn't near zero. The rule you quoted wasn't this. In your quote the operator could decide the draw/resign with looking at the score or consulting the team first. Nothing there that the team had to look at the score. How else could game 4 be a draw? > > >>I think this missing feature from the program should be brought out so any >>potential buyers know which behaviour he can expect from the program. > >The feature that is missing is "GM-style draws". But commercial Junior will most >likely have a "contempt factor" that the user can set as they wish, from "never >draw" to "always draw". Would not help as long as the engine gives wrong score. Else, this is a point that I have tried to point out when discussing the lack of draw messages in the uci-protocol, that this can be simulated with sending 0 score instead of drawscore. >>When it is known after the game that they not only have looked at the output >>from the program but also analysed the position before decide to offer/accept >>a draw, isn’t it then something with the ethical norm of playing a chess game >that is broken. > >I think you can't forbid to human spectators of a game to analyze the position >that arise : they will always do, even if only in their head. >So if you rely on an human for deciding draw offer, you can't avoid he analyzed >the position. Two quote from Mr. Ban: "Boris suggested 27. Ba3 Rc4 and we let DJ analyze on the backup machine. It would continue 28. a5 and felt quite good about itself." "After this move Boris, in his gentle way, became more insistent: 28. f4 f6 and the black king comes to e6. If white now plays a5, said Boris, then I as black play b5 and I want to win! We both could see on the analysis machine that Junior wanted to play a5 in every possible variation. I had to agree that Boris was making more sense than Deep Junior" This is analyzing and not viewing the screen. And what more is, they thought Junior would 'blunder' in the next few moves so they was afraid to lose because of this analysing. Beside from this. The team was part of the game, reported to the arbiter before game start, so they can't be compared with spectators. (Wonder if this backup machine was reported as a team member :) ) > > >>>I think this answer your question. More details here >>>e. When a draw offer is made by the human player, the machine's operator will >>>promptly enter the human player's move into the computer. The machine team may >>>start consultation at this point (off stage), viewing whatever is on the >>>machine's screen. >>> >> >>Ok, so it was booked that the was allowed to see Junior's output, but this is >>totally different from start to analyse the game on a second machine before a >>decision is made. > >The rules allowed them to read machine output, principal variation and such. The >fact that they probably couldn't see junior screen from the team room and >switched on the second machine for checking what junior was "thinking" doesn't >seem such a big deal to me. > :) Odd Gunnar
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.