Author: jonathan Baxter
Date: 19:45:33 11/11/98
Go up one level in this thread
On November 11, 1998 at 19:28:18, Robert Hyatt wrote: >On November 11, 1998 at 12:42:08, Howard Exner wrote: > >>Just read Shirov's article in Chess Life, Nov. 98. This player is >>becoming one of my favourites. He annotates two of his games and also has >>some comments on chess computers. Quoting Shirov, >> >>"Even though the games are completely different (one extremely positional in >>nature and the other very tctical), they had one thing in common; my >>difficulties in analysing them with Fritz 5! I entered some of the lines I >>calculated during the game into Fritz5 and it proved useless!" >> >>Well so far so good. He in a way is giving Fritz5 a plug by revealing >>to the world that he uses it. He also says that in the positions in his game >>Fritz5 was useless. That's ok with me also since all computer programs >>have a certain blindspot for particular positions. >> >>Shirov continues... >> >>"Yes, the computers can also see many more simple things than the human >>GMs do, but can they compete with the best human players in depth? I am not so >>sure." >> >>So I am in a sense still agreeing as many magficant human moves are >>difficult for humans to find. But now my inteest is peaked as to what the >>positions are. The tactical position is this: >> >>6r1/2rp1kpp/2qQp3/p3Pp1P/1pP2P2/1P2KP2/P5R1/6R1 w - - id "Shirov - Krasenkov"; >>bm Rxg7; >> >>This is a possible variation in the game and Shirov's comments are, >> >>"White would have a forced mate in 14 moves! Don't try to put this particular >>position on Fritz5 or any other program, as it would never suggest 33. Rxg7+! >>as the stongest move!" >> >>I looked at the diagram and thought to myself I'll bet that Rebel 10 >>would have a good shot at this as it seems to have a powerfull mate finder. >>Sure enough after 2:34 it plays Rxg7 with an ever increasing eval. That on >>an AMD K-6 233 in dos with 60 mb hash. >> >>So the moral of the story is not to generalize from one computer program to the >>next. Probably safer to say that "some computer programs will not solve >>such and such a position". My take is that the programs are as unique as the >>programmers who work on them. I think they are entitled that rather than >>lumping them all together. >> >>So Ed and other programmers whose program can solve this in a tournament time >>setting can email GM Shirov with the news that Fritz5 (or any single program) >>does not represent the entire pack. I'm sure some other programs will play >>Rxg7 also but I only own Rebel 10 at present so cannot verify this. What >>does the rest of the pack play? >> >>Just a small note that the point of my sharing this is not to belittle >>Fritz's play - it makes great moves in many different positions. It is more >>a sharing that all programs are unique in their strengths and weaknesses >>and that one program cannot possibly be singled out as representative of >>the whole. It would be as faulty to say, "Rebel 10 finds this position >>relatively quickly so the computer programs of today are routinely find >>such difficult moves." >> >>Enjoy the position. By the way, I'll enter in the other positional >>epd that Shirov was reffering to in a follow-up. > > >That's a point overlooked by most "non-computer-chess" types... If A can't >find this then no program can find it. Or if A can find it then all programs >can find it. Chess programs are a lot like people in this regard... all can >find some things, all will miss some things, but between these two extremes >there is a lot of room for one to find this, another finds that, etc... An observation that has led me to speculate in the past that the easiest way to win the WCCC might be to have a meta-program driving all the strong commercial (and free) programs. Maybe set Junior analysing the position directly and then have Fritz and Rebel and Crafty and .... analyzing the variations Junior throws out. Or maybe have a few analyzing directly and a few "verifying"... Jon
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.