Author: Uri Blass
Date: 03:54:07 03/14/04
Go up one level in this thread
On March 14, 2004 at 06:25:00, GeoffW wrote: >Hi > >I have been playing a few games with Fruit recently, the thing that I found most >surprising about it was the amazing time to depth of its search algorithm. My >progrma is not especially slow but it was lagging usually 1 or 2 ply in the >opening and 3 or so ply in the middlegame. >I am sure Fabien said it was a slow searcher in one of his posts !! I wish I >could make my program as slow a searcher as Fruit :-) Fabian talked about nodes per second and it is slow only relative to a fast searcher like Crafty. It is not slow relative to real slow searchers like tscp or Hiarcs or Biglion or chessmaster or Abrok or Diep or gothmog or chess system tal or Movei(I only mentioned these programs because I know that they are slow searchers and I do not care if the reason is not doing optimization(tscp case),Big evaluation(Diep's case) or expensive search rules(probably the case for part of the other programs like chessmaster or biglion) or different reasons in the case of movei. I do not know about the speed of your modified tscp in nodes per second and it may be also faster than your program. My reasons include not having function to generate only captures or only captures and checks when my qsearch generate list of legal moves before finding the captures and checks and changing it(at least in the qsearch) means also changing my evaluation because I use the number of legal moves for my mobility evaluation and I was afraid to change it because it is not a trivial change and I am afraid from bugs in the process. I am sure that a change in my mobility that is not going to count number moves can be productive both for faster search in nodes per second and for better evaluation(I should not evaluate backward moves the same as forward moves) but the problem is that I am less optimistic about my ability to do it without bugs so I even did not try to do it. Uri
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.