Author: Mike S.
Date: 19:42:58 01/23/02
Go up one level in this thread
>This code can prevent Crafty to capture the knight or the bishop when it is good >to do it and this is the reason that I do not like it. > >It may help to get better results against humans but I do not care about it. >I prefer to see code that help the computer to see faster if the sacrifice is >right or wrong. The logic behind that kind of anti-trojan method is the same as in every decision by *evaluation* IMO, when the program cannot verify it "just" by calculation alone. Calculation and evaluation cooperate always; a trojan sac is just an extreme example. I think it is ok to implement anti-trojan code, especially if it is a bit flexible (I didn't test that), because this is a piece of chess knowledge to be put into the program like many other parts of chess knowledge. Don't take on g5/g4 under certain circumstances. It may not be perfect, but much better than not to have it at all, when users which are generally much less in strength, can constantly beat it with that trick. I do not expect chess programs to re-invent the wheel by themselves :o) I prefer to have optimal strength by whatever means. Btw. easier ways to apply the trojan trick have been rejected by chess programs for some time in previous years, just because of the higher depths. Temporarily I thought this problem was solved, but it re-occured because the threat can be "masked" in a skillfull way. >If I play a correspondence game and my opponent play the trojan horse sacrifice >then I want a program to tell me if it is a good idea or not good idea to >capture and Crafty's code does not help here. There is an easy solution: Just play hxg5 on the board manually, and analyse from there. But OTOH I agree as far as I'd say it is useful to be able to switch off the anti-trojan algorithms, for more complicated situations when there is doubt that the progam may miss something in analysis. Regards, M.Scheidl
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.