Author: Vincent Diepeveen
Date: 15:10:22 07/19/99
Go up one level in this thread
On July 18, 1999 at 10:13:18, Robert Hyatt wrote: >On July 18, 1999 at 04:37:02, blass uri wrote: > >> >>On July 17, 1999 at 16:33:25, Aloisio Ponti Lopes wrote: >> >>>On July 16, 1999 at 23:18:36, Robert Hyatt wrote: >>> >>>>>the question is impossible to answer. To produce an Elo rating (which >>>>would imply FIDE to most of us) we would need ratings for at least the >>>>simple majority of the programs that participated. No such figures are >>>>available, since none have FIDE ratings... >>> >>>Well, Crafty is not listed in SSDF list (yes, I know the reason why). I think >>>Crafty is a wonderful program, it plays very well standard chess and at faster >>>controls it is really great. My question is for Dr. Hyatt: what ELO Rating do >>>you think your program would get if it participated in human FIDE tournaments >>>for one year? I would say at least 2550, do you agree? >>>A. Ponti >> >>I think that parallel crafty is not going to get 2550 in fide tournaments >>because >>crafty has one disadvantage(the code source of crafty is free) and if crafty is >>going to play in fide tournament the opponents can learn exactly what is the >>evaluation function of it and go to positions when the evaluation function of >>crafty is wrong. >> >>I think it does not happen in ICC because the good players do not take the games >>seriously and do not want to improve their rating(If GM's want to improve their >>rating they can play standard time control of 120 60 only against computers and >>they do not do it). >> >>I believe that not free programs of the same level of crafty in >>computer-computer games have better chances for this reason. >> >>Uri > >I think you are both right and wrong. There have _definitely_ been cases >of people studying the source to look for weaknesses. [a 2275 player raises his hand] >However, there are also many cases of GMs playing _real_ anti-computer >chess there, which gives many programs a lot of trouble. The biggest >shortcoming I see in commercial programs on ICC is predictability. They >want to play the best openings possible, which lets a few players find a >pattern they can exploit with anti-computer play... Variaty in book is indeed a big word, but strong players remember simply everything they ever played and studied, so then you have a major problem when playing outdated lines which are +- when human has white and -+ when human has black, as he knows how to handle it. Secondly i think that computer vs human will never give a good compare, as grandmaster are professional chessplayers who have specialized in beating humans, and not specialized in beating computers. When playing a computer it's like telling a child who is a bicycle champ to join a race contest with a formula 1 car against a very predictable shermantank. It's tough to pass a tank if you don't know how to steer a F1-car and if you don't know how to drive a f1 car (and that's tough for kids, cuz they have a lot to learn) then you're history. I'm still amazed at the internet in what ways however those kids manage to beat computers. Sometimes they are completely suicidal busy, and then suddenly they get a chance and grab it. Looking at % however then i'm quite positive about programs with a good preparement getting all 2600 FIDE easily. Above 2600 however it's gonna get tough, as those guys make very little mistakes. Mistakes as made by Sokolov against Fritz in paderborn demonstration game, is of course from the kids level. It was a direct and obvious crash at move 11 (of course it was hard to see Bh6 in advance, but the strategy chosen by Sokolov sucked bigtime), which kind of surprised me a lot. Sokolov did *anything* to get an endgame, and managed in that, as he had seen a game of Anand winning with it against fritz, but forgot the prize of it, which he could have seen at icc if he had given a little bitmore attention there to a another few tens of thousands of games played at the icc servers, nearly all lost by the human playing a similar position with the black side against the computer. And of course Anand can't be compared to Sokolov, so when i saw the opening i already had bad feelings for Sokolov, and when i entered the 11th move in DIEP and saw the score jump, i knew enuf. So in short: crafty makes a hell of a chance against GMs Some will try to draw a lost KRPKR endgame as "computer sucks in endgames, so easy for me to draw this". Others professional dudes will play after 1.e4,e5 2.Nf3,Nc6 3.d4 the move 3...d6, as "an endgame against computer can never be lost". Another few happy weirdos will try what i used to play against crafty; the already fixed 1.e4,e5 2.c4, as "crafty will *never* play f5 here but Nf6 instead and then also not get the idea to play f5 here any day, tip from Diepeveen, and it still works against genius, fritz and some other progs so it works against all comps". And the only one winning will be still Vaganian... ...when he has black.
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.