Author: Tord Romstad
Date: 11:29:16 08/02/04
Go up one level in this thread
On August 02, 2004 at 13:58:27, David B Weller wrote: >Hi all, > >GES is updated again, for those who care :) Hi David, This is the first time I heard about GES. I've already downloaded the source code, I hope I'll manage to compile it on my Mac! :-) >Somehow, I have managed to get a fail-high-first-move rate of 96%. However, when >I estimate branching factor by: (elapsed time iteration N / elapsed time >iteration N-1) and average over entire wacnew.epd, it shows 3.21 branching >factor. > > Isnt 3.21 high? I don't think running through WAC at 1 second/move is a good way to measure the effective branching factor. It seems plausible that the effective branching factor tends to be higher than usual in tactical positions, because the best move will often change very frequently in such positions. It is probably better to collect statistics from real games. I would also recommend you to only consider the numbers for the last two finished iterations at every search. In most engines, the effective branching factor is quite high in the first few iterations. > Isnt 96% for fail-high-first fairly good? Yes, it sounds good. > Is there another metric for move-ordering? Nothing that I am aware of. But I think tactical positions are not the right ones to use for move ordering statistics either. >wacnew results 255/300 1sec/pos 1.4 Ghz >av. dep nodes qnodes % fh1 branch >6.47333 45811K 11081K 19% 96% 3.21639 > >I dont think these stats are too stunning, but Ges will challenge Ruffian105 and >Gothmog sometimes :) I hope to see it challenge Gothmog on my Mac soon! :-) Tord
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.