Author: Thorsten Czub
Date: 16:14:06 04/07/98
Go up one level in this thread
Hm. Nice chat you have. I am no expert on CSTal but want to explain WHY we have done WHAT over the years. Chris is such an unbelievable guy, he edits the whole day into his big (really BIG) sources and has sent me at least ONE version a day ! I was only able to NUMBER them and to try to force him to save his sources the same way. Than I tried to find out (very fast if possible) what the difference has done to the engine. We started with our work a long time ago. It was Chess Champion 2175 or even earlier. Than Complete Chess System was done. It was nice but blundered within 4 plies or even oversaw 1 ply threads. From this very weak (but still genius) program we decided to start with concentrating mainly on the evaluations and implementing and changing the evaluation. This took time. The program grew stronger and stronger. The program played more and more speculative with very high evaluations. One day chris developed the TAL-function. From this time the positional playing type B program developed into a killer. It sacced pieces like hell, attacked wherever it was possible. If there was a bishop x h7 thing on board, it did it. You can imagine that it lost many games. We did not give up and tuned this function. Implemented even more attack knowledge and: simsalabim: suddenly the results were stable and we got nice versions. One main problem destroyed our work: the bugs ! The fact that chris often has not backupped his sources. Very often his changing while I was downloading the latest engine, confused mch. I had to look through. If you test ONE version a day, if you telephone with the programmer almost once or twice and exchange data, than he gets your brother, and the program gets your big love or baby. Suddenly you FEEL for it. Suddenly you can see without playing games if the version is a progress or not. Often of course you make mistakes and anything was wrong. For a long time we had 2 effects we were unable to handle: 1. the draw bug 2. the black results are weaker than white results. The draw bug is very strange. It is NOT there as long as cstal is fighting for a good position. It only destroyes the game when the opponent is dead. Suddenly, when the opponent feels itself -2 or -3 pawns the draw bug appears and ruins the position. It makes a draw or plays a complete stupid move. So we often had winning games lost in the "last minute". In the times we had 60% against all kinds of opponents (Moritz was unable to reproduce these results - I don't know why. I - one the other hand - was unable to reproduce his very good Fritz5 results. Strange...) ONLY the white score was 60%-. The black results were 30% or even less. Was it a bug ? The opening book ? Is black really that much weaker ? Or is tals style that difficult to play with black ??? We searched for the bug. dozens times chris claimed: I found the bug. He send the fixed version and suddenly it was the same problem. In the end we were unable to manage to fix anything. It is a good program with nice features. It comes on CD and the program plays succesful and sometimes very interesting. But the CD-version comes with a bug. I can live with this bug. Better a program that can generate winning position and sometimes makes a draw out of it than a program that is NOT able to get these positions. The "Paris" version (287c) was maybe fixed concerning the draw-"bug", but it seems to play weak. ONE reason I don't understand why Torsten Schoop uses it. IMO the CD-version plays stronger. After Paris (where I tried NOT to change the versions daily :-)) like chris would have liked to do) I was KO and also chris needed a holiday. In this process we had some time where our friendship was suddenly ruined. Somebody tried to cut our friendship by sending Chris faked emails that SHOULD be written by me. It never was. I never understood why chris changed into the Tueschen-club. I don't think I can live with a friend attacking my other friends (Dirk, Ed, Moritz, Peter...) so I lost relation to Chris W. and CStal. I have not much experience myself with cstwhite or cstblack. I don't know which version is stronger. Maybe we can find out together. Is it worth doing it when the author is not interested in us and only in fighting his nightmares against CCC and the rest of the world ? Chris told something that he will not anymore programm on chess system tal. Is the project now over ? IMO this would be sad since the program is really a surprise concerning way and strength it follows. It would be a nice thing to help chris to make it stronger. But it seems chris is more interesting in fighting against ENEMIES of all kinds (seems to be a virus he got from Rolf Tueschen, bad desease/infection)than in programming on cstal. I hope we could support him continuing his project. Maybe if you request enough, he will concentrate on programming instead of fighting wars against Bob/Bruce and all the others he calls with mean names. I don't like this too but I cannot change it anymore. I lost control over it. So far from the CSTal frontier. In the moment it looks CSTal is OVER and there are some guys working on tranporting the user interface into a win95 stuff, but it seems CHRIS has stopped working on the chess-engine. A pity. If you want I can put cstwhite and cstblack on my autoplayer against several opponents and we can follow how the results develop. I would not use time controls smaller 60" per move. Till later. It is a pleasure to talk with other people about a topic I was talking alone with chris for at least a few years. It makes fun to share. I hope you join it as much as I have done it.
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.