Author: David Rasmussen
Date: 03:02:45 12/05/01
Go up one level in this thread
On December 05, 2001 at 05:15:56, Sune Fischer wrote: >On December 05, 2001 at 03:24:29, David Rasmussen wrote: > >>You don't seem to want to comment on the pawn hash collisions in Crafty, that I >>have mentioned. Is that because they don't matter or because you don't believe >>my experiment and think I have a bug? >> >>/David > >Who are you asking? Sorry, I am asking Bob. >Whether or not 32 bit is sufficient depends on which "resolution" you require. >If you need to distinguish between millions of different positions, then 32 bit >will create too many collisions, probably somewhere near the numbers you've been >talking. But if you "only" have a few thousand different positions (which is >typical if you hash pawns _only_) then on average you should have about 1% >chance of a collision in an entire game. > >Try and count how many different positions you get, this is really the key >factor. > First of all, we're both talking about hashing pawns only. And that is what I have been talking about all along. Now: I disagree. They key factor is how many actual collisions returning wrong scores (and other wrong data from the pawn hash table) you have, regardless of how many positions etc. I know that there aren't that many pawn positions during the cause of a game, but I really don't care. The important thing is to measure how often the hashtable adds incorrectness that _would_ not have been there, had it not been for the pawn hashtable, i.e. a defect or, if you will, a bug of the pawn hashtable. And whether this rate is "high" or "low", I don't know. Until I am sure it is not "too high", I will stick with a rate of zero. Your argument that there is about 1% chance of collision per game is all well and good. It just doesn't matter, since Crafty have more than 1 collision a second with appx. 150 kn/s from the initial position. I am asking several questions: Is this expected or a bug? Is it too high? If not, why? /David
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.