Author: James Robertson
Date: 13:45:54 09/04/99
Go up one level in this thread
On September 04, 1999 at 02:46:52, Terry Ripple wrote: >On September 04, 1999 at 01:21:20, James Robertson wrote: > >>On September 04, 1999 at 01:03:26, Terry Ripple wrote: >> >>> >>>When i run engine vrs engine matches i set it that each engine plays white and >>>black in each opening, this way it totally makes it more fair! >>> >>>Regards,Terry >> >>I think not. It _will_ skew the results. >> >>For instance, say you pick the King's Indian and make both programs play it. It >>is a very hard opening for a program to play because black accepts a big space >>deficit in return for a kingside attack. Most (all?) programs don't understand >>this and as a result white has a _big_ advantage. So if you play two games >>between two engines, white has a good chance of winning both, even if one engine >>is a lot stronger than the other. Ditto for all Gambits and even a lot of other >>openings that 'seem' equal (and maybe for humans are). >> >>Maybe you pick a very tactical opening, or an opening that revolves around >>kingside attack. Some engines are worse tactically than others, but rely on >>their book to avoid tactical positions. This time, instead of the games being >>skewed in favor of a certain color, they are skewed in favor of a certain >>engine. >> >>My program's book is tailored to its needs, and fits its style. To separate the >>two will result in a _big_ decrease in playing strength. >> >>James >--------- > >Hi James, > I thought about this and i believe that you have a very good point which i >never really thought about! So what you are saying is if engine A should choose >e4 and then engine B chooses d5 (Chosen frequently by CM6000)and it now leads >into the Center Counter which CM6000 likes to play and now the next game i >force engine A with black to play that exact line of the Center Counter, then >this might not be tailored toward its style of play like CM6000 likes! Yeah.... my engine has a very hard time with black if it plays the center counter. Such a hard time in fact, I changed its book to never allow it to play the opening! > I thought the top chess engines are "tactical monsters" and that a chess game >is won by 99% tactics! So, are you saying that not all programmers are selecting >their opening books based on this because they programmed a style in to their >program to play more positionaly and choose closed lines of play rather than the >open tactical lines, perhaps because their program is not as strong a tactical >player? Yeah.... all programs are excellent tacticians, so when playing humans a game with as much tactics as possible is good. Even if a program doesn't play certain tactical positions as well as other programs, they are so much better than humans it is worth it to play the opening anyway. Against other programs, they are on much more even ground tactically, so choosing openings the program likes is much more important. > I have been playing a match between Hiarcs 7.32 and Crafty 16.13 (Single CPU) >with a time control of 40/120 + 30/60. The match will randomly select a opening >and then it will play that same opening again with colors reversed. I'am using >the opening book of Hiarcs for both. Now i know that Hiarcs has more position >sense than Crafty but the opening book don't seem to be hindering Crafty in the >least because it's is doing quite well so far in the match. It is possible that an opening that favors a certain color (that an engine choosing its own opening from the book would not choose) will be randomly chosen, in which case the better engine will be hurt.... Hope I make sense! Anyway, I am very happy someone is finally testing Crafty against the best engines at 40/2. If Hiarcs uses the Hiarcs book and Crafty uses the Crafty book on its ftp site, then both engines choose their favorite openings and you have a completely fair single-computer match. :) James >They played a total >of 8 games so far with a score favoring Hiarcs 5 to 3. I know this is not nearly >enough games to make an accurate judgement but it's a start and i'am not a >rating agency trying to prove who's the best but only doing this for the >enjoyment of playing over the games afterward and posting here for others who >might enjoy the games. You can still enjoy some beautiful moves from the engine >matches, even if they aren't optimized for their maximum strength! > >Best regards, >Terry
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.