Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Jonny-Movei results in the nunn match on AMD3000 ponder off

Author: Jorge Pichard

Date: 08:11:27 07/23/04

Go up one level in this thread


On July 23, 2004 at 06:37:58, Uri Blass wrote:

>On July 23, 2004 at 06:19:02, Joachim Rang wrote:
>
>>On July 22, 2004 at 12:56:49, Uri Blass wrote:
>>
>>>Conditions:
>>>
>>>Jonny 256 mbytes hash
>>>Movei 64 mbytes hash(I suspect that more than it is counter productive for all
>>>time controls and even 64 mbytes is counterproductive for fast time control
>>>except very fast time control of 3 minutes per game or faster because in very
>>>fast time control movei does not waste time on clearing the hash because it
>>>assumes that it has no time for it and the old entries in the hash are still
>>>productive for better order of moves and movei does not suffer much from useless
>>>entries)
>>>
>>>Hardware A64 3000(2ghz)
>>>
>>>Movei-Jonny 36.5-13.5  at 1  minutes per game
>>>Movei-Jonny 34.5-15.5  at 2  minutes per game
>>>Movei-Jonny 33.5-16.5  at 3  minutes per game
>>>Movei-Jonny 27-23      at 4  minutes per game
>>>Movei-Jonny 23.5-26.5  at 10 minutes per game
>>>Movei-Jonny 2.5-3.5    at 20 minutes per game
>>>
>>>games played under winboard.
>>>It seems that Movei is simply weaker at long time control and it is not
>>>surprising because it has poor order of moves and bad use of hash tables.
>>>
>>>The surprise that I have so far is the good results of it in WBEC.
>>>
>>>Uri
>>
>>
>>what is a bit disturbing it, that you gave different Hash to different
>>opponents. As you wrote 256 MB is more than enough for a 1 0 game and perhaps
>>even hurts Jonny in the fast games (just a speculation).
>>
>>regards Joachim
>
>Based on watching some games it did not seem to me that big hash hurted Jonny(it
>is not a program that clears it's hash after it's moves and I believe that it
>did not lose games on time but I will check later).


I used 64 Mb of hash for Jonny at blitz and it did great at Blitz, but at longer
time control game in 5 hours I'm giving Jonny 256 MB.


Jorge
>
>I think that the rule should be for most programs that bigger hash is never
>worse except programs that clear the hash between moves(It is the case for
>crazybishop based on asking here but I did not ask about Jonny before starting
>the match).
>
>practically it seems that Jonny used less than 256 mbytes but more than 200
>Mbytes and maybe it cannot use exactly 256 mbytes(I used configuration for 256
>mbytes that comes with the engine).
>
>The result in the match at 20 minutes per game was 9-9 the last time that I
>looked and now I am not near the computer that is doing the match and even in
>case that I were near the computer the only way to check the results could be by
>looking at the pgn of the games(I simply followed every game from 3.5-2.5 for
>Jonny to 9-9 so I did not need to look at the pgn at that time).
>
>I expect the match(20 minutes per game) to be finished in a few hours and I will
>report the results.
>
>Uri



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.