Author: Thorsten Czub
Date: 15:17:39 09/16/02
Go up one level in this thread
On September 16, 2002 at 17:38:42, Rolf Tueschen wrote: >I'm a bit astonished by such a tone here, Dann. :-))) >To answer your question I remind you of the point that I originally _thought_ >that Thorsten posted or quoted the games, because they should "prove" or show >that the created new style played "stronger" than the default version. right. you play autoplayer games . yuo count the result. and when the result of the match is better, you have a stronger version. thats the ssdf-way. they do not value the HOW. since i valued the HOW, i knew the style is better. only i had to play autoplayer games to show that. the games have been published. you - unexperienced with autoplayer games (you have only 1 pc, right?) see the games and wonder about the openings. but thats all normal. so the problem is not the games, but YOUR relation to the information you see. > My >question was if such games, my 5 examples, could do that successfully. your examples show IMO nothing. all they show is how senseless it is to measure strength by autoplayer games. > I thought >'No'! If you doubt my original opinion then why these games were posted at all? WHY these games were played ?? why do you buy a chess program when you do not understand about computerchess ?? play with it. but do not post here. this is not a chess forum but a computerchess forum. the games have been played to prove that macheide is stronger then rebel.eng. you do this with an event. you make games. i do this with 2 pairs of machines. 4 machines run day and night 40/120 level and produce data. this data is collected. the data is ordered, posted and there. in the same way XP rebel.eng played 10 games and lost 1-9 against fritz. style 51 will not lose that high. so it is stronger. logical, or ? now it could be 51 is a killer for fritz. therefore you try against all kind of other programs. again you play games . day and night. the machines do that almost automatically. in the end you count. thats all. thats the way it works. we have done this way with Chess System Tal (Chris had 4 autoplayer pairs in his company. One against genius, one against mchess, one against fritz ...). but each new version of program needs tuning. so you have to tune the parameters. thats a very difficult job. you have to find out how to set the weigthings. this is been done in the stage from version 1 of the style to version 56 of the style (in my case). in the end the program is tuned. >They stood in context of Thorsten's claim that his style was stronger against >Shredder ans so on. But the games don't prove it. the games prove it. >I found that the games contained a cook against Shredder. Explanation: because >the Shredder book contained the weak line. What should this prove about the >Macheide style??? Nothing IMO. it should prove nothing in this game. but you do not play ONE game. you get it: you do NOT play ONE game. you make MORE than 1 game. >I asked a scientifically interesting (for me!!) question how people, I meant >all, could detect such cookes games in testing. Because it made the tests weaker >by definition! (Would you now doubt my scientific interest I showed from my >first posting on? Either about SSDF or the DB2 team.) opening books are part of the programs. or how do you think Alexander Kure, Jeroen Noomen or Vincent or Sandro spent their time here ? just for reading your great posts ? they work all day and night to find a book line and a position that is good for their programs. >They are cooked because of the opening book. And I thought I had made a valuable >_discovery_. ???? but its obvious. any chess player (who is capable to play chess !) sees this instantly when replaying the games, or ? this is your discovery ?? oh man. >Just by playing through the lines. I found the examples. right. how else should it go. by asking nostradamus ? >Please do not lay bad intentions into my mouth when they definitely are not >there! nono. why should we ever lay bad intentions in your questions of a newbie. its all only a question. >Of course I thought that it was a bad thing for Thorsten to post these games, >but I did never even _insinuate_ that he tried to _cheat_ with them. For what >purpose should he have done it??? the reason, you scientific expert is , to present the data. you cannot leave out the games only because the opening was bad. if you leave them out, it would suggest a DIFFERENT judgement.
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.