Author: Thorsten Czub
Date: 23:50:28 03/16/04
Go up one level in this thread
On March 16, 2004 at 16:25:27, Uri Blass wrote: >I do not understand. >What is the reason for his objection that somebody else will continue his >program? > >If he has objection then what is the reason that he gave you his source code or >maybe I understood wrong and you were joking and he did not give you the source >code. > >Uri uri that is a long story. in working on CSTal chris send me nearly 1 version a day. often he changed something and later did not know what it was. i then made the offer to send me not only the versions but also the source code of this verison, this way we would be able to start from this verison if the change was any good. the source code is btw. very funny because it has many funny comments. i don't think that chris wants to do programming again, and i also think that he does not want to see anything he had done continued. he left the scene in anger. he even wanted that i should also leave computerchess. but i did computerchess before i worked with C.W. and i will do it after him. our relationship reminded me on this "odd couple" with felix unger and oscar. you maybe have seen this american serial with walter matthau and jack lemmon, that was a stage-play. chris was oscar, and i seem to be felix trying to keep anything save and clean :-)) my job was to find the bugs, to make sure they get fixed, to make to-do-lists for the programmers, to make sure the project makes progress. we had 4 or 5 autoplayer systems in his small company. there we tested the verisons and learned about the changes. it was a fight with the programmers (who often had no idea about chess or chess GUI's) and with chris lazyness and the sloppy (is this the right word when somebody is not accurate) way of editing the code. on tournament i had to make sure the BEST version plays because chris liked to change the engine with an UNKNOWN engine he just edited a few moments ago :-)) in the end we all made progress. it was even planned to do a version 3 with a rewritten engine, in version 2.5 you can see those efforts. but then chris sold anything and had big arguments with the scene. for chris computerchess is not his topic anymore. for me still it is my hobby. anybody has the chance to write his own kind of CSTal just by trying with a chess program that is not always playing ACCURATE. the idea in CSTal was to have a chess program that creates ideas. for this, you need an inaccurate part. the strength of CSTal came IMO mainly by chris evaluation functions. he had a kind of formulas, he himself created, from his experiences as a chess player. weird stuff. generalisations. but it worked. chris tried to use teaching chess books and videos to translate them 1:1 into the program. this way CSTal knew a lot of chess stuff, and used it for dreaming about a plan. the job of the search was to make sure these ideas get realized without the opponent be able to destroy it. one idea was that the opponent is working ACCURATE, so if we pull him into FOG HE has to calculate accurate for a defense, not CSTal. so we thought that in fog, CSTal is better because of the knowledge and the generalisations. e..g in a game against Dark Thought at the championship in paris CSTal knew that pawns need the KING to be strong. this easy generalisation helped CSTal by cutting the King of Dark Thought to the edge, cutting it from the pawns. Dark Thought came very deep, ran on a DEC alpha experimental version. but it was dump. IMO the strength of CSTal came from the mixture of risky ideas and experience chris tried out because he was a chessplayer and a little insane :-)) IMO cstal works other way arround than genius. genius calculates 1-3-5-7-9 plies very selective and 2-4-6-8 plies nearly brute-force. this way it sees all threads while it is NOT ALWAYS seeing a winning key move for himself. the result is a playing style that is defending, boring, not really doing something, waiting. CSTal is not computing all moves of the opponent. it prunes much away. this way it tries out many things and believes it would work. this results in a nice playing style.
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.