Author: Ed Schröder
Date: 12:24:36 05/30/00
Go up one level in this thread
On May 30, 2000 at 00:28:47, Robert Hyatt wrote: >On May 28, 2000 at 16:37:32, Gian-Carlo Pascutto wrote: > >>On May 28, 2000 at 10:02:05, Georg v. Zimmermann wrote: >> >>>From my tests it shows that it sticks with the hash-move about 50% of the time. >>>Should this number be higher ? >> >>Hmm...if this number is also effectively your 'move ordering percentage', >>which I assume it is, it is quite low. I'd expect it to be at least about 75%. >> >>> > > > >The classic definition of a "strongly-ordered tree" is this: If, for every >node where you fail high, you fail high on the first move at least 90% of the >time, then your move ordering is good." If you are much below 90% and already >have a serious problem that is not hard to fix. The traditional ordering ideas >holds Crafty at 92% and better for most of the game. I can't understand the 92%. A perfect mini-max search requires many many nodes an alpha-beta cutoff will not work and you are forced to search all the nodes of the ply in question. And this number is certainly much higher than 8%. The way I count is to calculate the percentage around the alpha-beta check. I get 65 to 70% and this number is pretty stable for each position I tested. So the question is how you count, and your interpretation of a fail-high. When I read fail-high I interprete this as a normal alpha-beta cutoff check. Ed >>>I was very dissapointed when I didn't notice any speedup after my changes. What >>>speedup should I expect ? Something like 0.5-1% or more like 1ply ? >>> >>>Am I doing something wrong or does this simply not matter as much as I thought ? >> >>It's strange. I did the same thing in the past and arrived at the same result. >>My eval is pretty small, my movegen is very expensive yet my speed was hardly >>any better. I didn't get it. I still don't. >> >>I still don't get how some programs run 10x as fast as mine either.(Little >>Goliath). What the hell do they do to make those things so darned fast? >> >>-- >>GCP > > >Easy: Never do now, that which you can defer doing until later. Because if you >defer doing it, a cutoff will mean you don't do it at all.
This page took 0.02 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.