Author: Robert Hyatt
Date: 17:04:18 11/17/03
Go up one level in this thread
On November 17, 2003 at 18:03:50, Daniel Clausen wrote: >On November 17, 2003 at 17:52:05, Robert Hyatt wrote: > >>On November 17, 2003 at 17:35:42, Daniel Clausen wrote: >> >>>While one should be careful when testing engines with only a few positions, I >>>wouldn't be surprised if Crafty would have the very good (if not the best among >>>computers) chances against humans in a similar match. The effort to concentrate >>>very much on games against humans has to pay off after that time. Too bad that a >>>match GK-Crafty is not realistic due to commercial constraints. >>> >>>Sargon >> >> >>I would hesitate to draw conclusions from a couple of positions. The >>entire game has to be played. I certainly feel that the commercial >>programs are better-tuned overall, even though Crafty will clearly have >>a right idea here and there that others don't.. > >I know that a couple of positions don't mean a lot. It's more a feeling than >anything else. :) > >About being better-tuned overall: don't you think this is more important in >comp-comp games than in human-comp games? Again, I can't really prove it, but I >think that in a human-comp game it's more important to have certain eval-terms >than having them well-tuned. What do others think? Any "proof" for one or the >other side? > >Sargon OK... a couple of cases. In comp vs comp, tuning is _highly_ important. In comp vs regular humans (non-super-GM, etc) then it is probably more important to have a sophisticated piece of knowledge than to have the knowledge weighted accurately. But against Kasparov, I believe that accuracy is just as critical as the knowledge itself. Because he give you so many things to choose from, and if you choose wrong, you lose...
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.