Author: Dave Gomboc
Date: 18:54:21 05/11/99
Go up one level in this thread
On May 11, 1999 at 20:42:30, Thorsten Czub wrote: >branching factor: > >my experience is, that we have to INCREASE the branching factor ! >example: > >without nullmove on cstal gets arround 3-4 branching factor. >with nullmove cstal gets normally <3 , often 2.5 or even less. > >but : imo this does not help, the opponents use nullmove, cstal uses >null move, so how do you want to defeat your opponent, if you cannot >see more than your opponent does ? >IMO alpha-beta and nullmove are what scientist have tried in early days: >they tried to come near to the sero-point in temperature! > > >but this does IMO not help us to play chess. >pruning has not the target that all programs behave equal and >have similar mainlines and play the same moves and come to the same >conclusions and get stronger and stronger each version and day (incest!!), >thats NOT the target: the target should be the opposite: >more variance... Variance to take your opponent into a position they do not understand sounds like a good strategy to me. Variance just for the sake of playing a second-best move isn't so good, though. Telling the difference between the two may or may not be a lot of work. >imo a NEW chess program should NOT follow the old way. it should go a different >way. and it should NOT optimize itself by doing what the others do. >because this cannot work. It can work if you do it better. :) Of course, going a different way might work too. >so - branching factor is only important if your chess strength comes >via search. than it is maybe important to have an efficient search >and to decrease branching factor. Okay. >maybe it is important for all the dump programs who have nothing than search >to generate strength. but - only for them. And the open-source examples of something different that plays well are where...? Dave
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.