Author: Thorsten Czub
Date: 15:22:54 02/18/00
Go up one level in this thread
On February 18, 2000 at 14:25:23, James Robertson wrote: >I don't know. I have seen CSTal in three tournaments; in one it finished dead >last among comercial programs, which program. i remember that wiesenecker made a tournament. but he used the wrong style. the same style harald first got his 0.5-9.5 loss. no wonder wiesenecker got weak results. when a style is so heavily influencing the programs evaluation that it plays 0.5-9.5 happens, you cannot imagine this would result in a good ranking in a swiss tournament either. in paris championship for a couple of rounds, do not remember exactly for how long, but it happened - you can research if you don't believe ME - cstal was in the leading group. i remember paderborn championship where cstal also did a good job playing nice games and making points. these tournaments were live and without any "cheating" or "tuning" chance. i remember david levy, or was it marsland, asking : why is cstal suddenly that strong? they meant of course : relative to the strength predecessors were. i took this as a compliment that the work chris had done was on the right way. not more. IMO cstal was always a litte handicapped because the executable was SO big that it did not fit well in the cache of cpu's. other engines did fit in the cache. when cstal is running on my k6-3 it for the first time fit's in the cache (the parts of the executable in charge for building the tree and doing the search fit in it) and IMO cstal gets a push in strength. again: on the same speed machines, but with less internal cache, cstal is 2 times slower. this is remarkable and i don't know any other program that runs 2 times faster on a k6-3 than on a k6-2 having same speed. i remember cstal-dos (or was it the win-version, don't remember it excatly) started well in another tournament i had done, and lost points in the last rounds against the very best (top). i think it was the same in paris. i think it was the same in paderborn (remember that in paris it lost due to strange bugs, remember that in paderborn it also lost due to strange bugs [before mating, it underpromoted into a knight! which lost the game!]) and i think in the moment there are not so many bugs in the 2.03 version. so : the cpu is best, the bugs are out, and cstal begins to fight like it did in paris / paderborn, and like in a tournament i did before (on 200 Mhz machines in those times). IMO chris should be able to increase strength if he would continue to program cstal. but he is doing shogi. he is banned here. and he is not interested in main-stream-adaption. i am sure he is interested in optimizing his knowledge-ideas about the paradigm-change. i think he is talented, whatever you believe about him, and in a certain way (i told this once , but i will repeat since it shows part of MY values too) he is continuing the ideas of botwinnik (pionier) and also david levy (mark V/philidor VI), thomas nitsche+elmar henne (Mephisto III), marty hirsch etc. pp. it would be a pity if he - like others (all the above people) - would be kicked out the society (of programmers) just because OTHERS believe his ideas are wired and he is an idiot. I remember that people said so about botwinnik. i remember too that botwinnik was kicked out by the fast-searchers and the mass-market ideologysts. nitsche was kicked out because people believed (the same fast searcher ideologists BTW) he is crazy and will never succeed or stand. he DID mephisto III and even won championship with it, a program doing 3 NPS on 8 Mhz 8-bit CPU and up to 15 NPS on 12 Mhz 68000 hardware. he was kicked out. by the money makers, the fast-search programs, by programs like superconny and Mephisto Brute-force-modules (A-strategy). levy had a very intelligent Intelligent-chess-computer, wooden and with cassette-tape, helmut pfleger original-voice commenting games, tv-output and video-board interface. you were able to save the games on tape! later he made mark v, b-strategy and later philidor VI. levy was kicked out by the fast-searchers ! mephistoI and later mephistoII and others were better than mark v. remember that mephisto II 6.1 Mhz played between 7-3 or even 9.5-0.5 against mark v. but mark v was a very intelligent machine IMO. but - again - the market - has no grace for intelligent solutions. they are kicked out. marty hirsch has a very intelligent and clever chess program. but it gets kicked out. because the user-interface is dos and old-fashioned, and the fast-searchers mainly overtake the slow-searchers. i don't think this happens because the idea is better. when things win, it has not often much to do with the idea beeing better. it often has to do with something else. guess chris will be or is kicked out for the same reasons. there is always a main-stream. and the main-stream kills the ones saying: hey - the direction you follow is wrong! whatever the target is, the people walk. when there is somebody saying: you have to stop this thing, it is shit. turn arround and try this, the people cruxify the guy. and they do it even more, the more right this guy is. not only in computerchess, also in politics or whatever field you can imagine. societies kill individuals when they are against the main-stream. they throw these people out of society. cruxify them. jesus. martin-luther king. olaf palme. kennedy. many of the black-panther organisation. the german rote-armee-fraktion (ulrike marie meinhof). rosa luxemburg. giordano bruno. whenever somebody is different, groups have problems to integrate. when pressure arises, they will put it on the "different" people. > in Thorsten's it is first, and in this match it >didn't do so well. :\ shredder is too strong for cstal. it outsearches cstal very often. has very deep search depths. 14 15 searches. even deeper than fritz. >The first commercial program Insomniac ever beat was CStal 2.03 (on FICS; >Insomniac actual had weaker hardware). And ? cstal is a kind of program that can win against all, and can lose against all. so it is. it is not dumb brute-force machine killing all weaker programs and having programs to play chess. > My overall opinion is that it can be >dangerous, but is quite a bit weaker than the top commercial programs. we are discussing how big this "quite a bit" is in fact. when my results are exceptional, why has cstal doing well in paris/paderborn, when it was live ? good luck ? random? >I am very curious to see more matches with CStal though. Thorsten's tournament >indicates that maybe it is a bit stronger than other evidence suggests? when you relate results (very important because you relate quantity) you must make sure that you don't relate apples with bananas. people testing on 2x k6-2/450 will get weaker results for cstal than people testing on 2x k6-3/450. joern gronemann e.g. tests k6-2 450 and gets weaker results than i do have, etc. etc. be careful when you relate quantities that the things you relate are really same conditions. jus my opinion about it. harald was right to play with fast machines. but he was IMO wrong to give shredder4 a 200. 2 x athlon500 would be better IMO. we would be more able to relate... but now we can only guess. again. whatever. i would like to see more people join chris', nitsche's , botwinniks, levy's, hirsch's and and and ideas. they are IMO worth to follow. even when success does not come over night. and you will in the first times lose against the fast searchers. this will change the longer you work. you should of course better not die before you succeed. otherwise people like F.F. will publish that you were mistaken, and you cannot defend from heaven by throwing asteroids down to earth. >James
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.