Author: Volker Böhm
Date: 15:11:23 09/22/04
Go up one level in this thread
On September 22, 2004 at 14:05:55, Uri Blass wrote: >On September 22, 2004 at 13:48:00, Volker Böhm wrote: > >>Hi Uri, >> >>as Spike is still only a patzer it needs a good opening book. Spike is a little >>weak in opening positions. It does not develop its pieces well if not done by >>the book. We are working on this issue :-) >> >>By the way: I don´t think that 20-30 games prove anything. I recently had a >>funny effect: I made spike a little faster (about 7%) just to see it getting >>worst. In a fixed tournament situation (against one other engine, 100 predefined >>positions, book = off) it dropped form 51% to 42% in 200 games. The engine is >>doing exactly the same, just a little faster (I compare node count in a WM Test >>suite with fixed search depth to prove it does exactly the same). >> >>Greetings Volker > >It is not a proof. >It is possible that the bug is related only to games and not to test positions. > >In order to prove that it does exactly the same you need to prove that you can >reproduce the same games with old spike with slighly slower time control. > >Uri I don´t see the difference between test positions and real game here - I don´t test playing strength. In test positions there are many positions generated that will test all sort of different position (my test has about 150 million nodes). If the amount of nodes stays the same then the program calculates the same. Theoretical it could go wrong but the number is high enough. To reproduce a game is not possible. Even a slighty slower computer could produce a different game. You have to search with fixed depth to be exact. Greetings Volker
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.