Author: Tom Likens
Date: 16:20:46 02/06/03
Go up one level in this thread
On February 06, 2003 at 18:39:36, Russell Reagan wrote: >On February 06, 2003 at 16:02:53, Tom Likens wrote: > >>If you graph strength vs. knowledge you will get a curve that >>will initially rise, but will eventually start to fall just for the >>reason you mention. As the speed cost of computing the newly added >>knowledge outweighs the benefit, the program will actually start to >>play worse. > >I think one important point here is that if you are just using evaluation for >evaluation, then the above is true, and you can only add so much evaluation >before it starts to hurt you. On the other hand, if you use that evaluation for >evaluation, AND for (say) pruning or reductions, then you won't have slowed your >program down, and you can theoretically pile on the knowledge and not lose any >speed (I'm sure this breaks down at SOME point). At some point (if you keep >adding knowledge), your evaluation should be good enough to start trusting it a >little more, and to do some pruning based upon it (or at least some reduced >searches). The question is whether or not you can reach that point where you can >trust your evaluation enough to accept the risk involved in pruning away >something important. Hello Russell, Don't most people do this already if they use a plain-vanilla quiescence search? I say this because one of the first checks performed by most programs, is to call the evaluation function near the top of the quiescience search and return if the stand-pat value is larger than or equal to beta- effectively forward pruning the branch. By the way, I believe it is this pruning, which is guided by the quality of the evaluation function, which provides the lion share of the "tactical" benefits of the evaluation function. I'm guessing that you mean the use of the evaluation function as a pruning agent in other parts of the tree. regards, --tom
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.