Author: Aaron Tay
Date: 10:13:25 04/07/01
Go up one level in this thread
On April 07, 2001 at 06:37:01, Vine Smith wrote: >On April 06, 2001 at 23:49:58, Robert Hyatt wrote: > >>On April 06, 2001 at 16:13:11, Vine Smith wrote: >> >>>I would like to get the "immensely aggressive" version of Crafty! This would be excellent for me, or any human opponent, to train against. I would be willing to work on a special opening book for such a version, if this were really a possibility. I think it would be interesting for programmers to try create a "Gambit" version of their program. Tom King's "Mad" is a really interesting example, I think currently Mad 0.04 released which he claims to be 60 elo better than the normal 0.90 at long time controls.. Of course it's no where as strong as Tiger, but I think it's interesting that the Mad version seems to be stronger than the normal one .. I;m currently, watching a nice game where Mad is outplaying Goliath , with a nice King side attack..I suspect though Mad does better as White, than black in many cases.. Perhaps we can have a "Gambit/Mad" version of Comet?? >> >>If you look at the documentation file, "evaluation" command, you can scale >>king safety up, make it asymmetric in the opposite direction that it is now >>so that the opponent's king safety is more important that crafty's king safety, >>then ramp up the 'tropism' term... >I'll give it a try. Are you interested in hearing of any results or testing? >Also, if you have any numeric guidelines, such as "Don't set king safety higher >than x, or it just gets ridiculous", I'd appreciate the info. I doubt it will work. My guess is you really have to work on adding more knowledge rather than just shifting the weights on existing postional factors.. But what do I know..??
This page took 0.01 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.