Author: James Swafford
Date: 18:39:30 07/31/02
Go up one level in this thread
On July 31, 2002 at 00:06:19, Jon Dart wrote: >On July 30, 2002 at 22:43:36, James Swafford wrote: >> Knightcap was strong, but it's >>definitely not in the top tier. >> >>Maybe Tridgell/Baxter quit to soon, and Knightcap really could've been >>a top tier program. Or maybe the reason nobody is using TD is because >>it's impractical for the large number of parameters required to be >>competitive in chess. Or maybe Schaeffer was right, and the commercial >>guys just aren't taking TD seriously. > >KnightCap wasn't bad at all. It was interesting to me that, not only >did they not believe in manual eval tuning, they also didn't believe >in manually constructed opening books. All the opening stuff was >learned, too. Right... I think there's something to that, actually. At the very least you don't have to worry about coming out of book in a blundered position (unless you blundered it the first time). > >It's an interesting approach. It's probably better than having an amateur like >me tweak things by hand, using a little bit of chess knowledge and intuition. >But if you have a GM to help you out (as Roman and others have done for Bob), But that's the point Jon! Most of us _don't_ have a GM to help us out. And I'm a patzer, so I need some way of tuning that doesn't involve me "guessing" weights. -- James >then I think you can make quite a bit of progress - these people have their own >finely tuned eval function that has terms you might not even think to include >(and if it's not included, its proper value cannot be learned). Next best thing >to actually having GM advice is at least to have strong players match your >program and demonstrate where it is weak - which is quite possible on ICC. > >I also think opening experts can add value, in a similar way. > >--Jon
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.