Author: José Carlos
Date: 12:44:13 02/08/06
Go up one level in this thread
On February 08, 2006 at 13:15:28, Dann Corbit wrote: >On February 08, 2006 at 04:43:48, Uri Blass wrote: >[snip] >>I agree until here except one thing. >> >>I do not know nothing about the real nps of rybka except the fact that they are >>manipulated so I do not know if it is relatively fast in nodes per second. >> >>I only know that it does not report correct number. >> >>> >>>It's clear Rybka relies much more on search than eval. >> >>Here I do not agree and I think that rybka has both good search and good >>evaluation. >> >>You cannot be at the top without good evaluation. > >I do not think we can say anything about Rybka NPS reporting until Vasik Rajlich >gives us his working definition of a node. > >I would be pretty astonished if he is intentionally giving us bad numbers. > >I see two possibilities: >1. He has a bug in his node computation. >2. He counts nodes differently than other people do. > >I think a conspiracy of false reporting has a probablility of about .000001% 3. (wild guess) he shows the nodes searched in the current iteration only. This would make technically possible to reduce the number in next iteration given agressive pruning and hash table effects. José C.
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.