Author: Christophe Theron
Date: 09:28:38 01/02/00
Go up one level in this thread
On January 02, 2000 at 06:29:43, Vincent Diepeveen wrote:
>On January 02, 2000 at 05:43:52, blass uri wrote:
>
>>On January 02, 2000 at 05:02:45, ed wrote:
>>
>>>Hi all,
>>>
>>>Am wondering if there is any benchmark for which CPU tiger favors. Also, is
>>>there a big diff between a K6-2 and K6-3 at same mhz for tiger? (for rebel I
>>>believe the speedup is 20%.)
>>
>>I read that rebel like the pentiumIII more than tiger.
>>
>>>
>>>Is tiger considered a knowledge pgm like rebel?
>>
>>The definition of knowledge is not clear to me.
>>I often find that fast searchers have better positional understanding.
>
>I am reading here the biggest crap out of history.
Probably the crap is not in Uri's text but somewhere between your screen and
your brain.
He is right. By searching fast a program can afford to avoid using a too hard
selectivity that prunes away good positional moves.
This has to do with "search being faithful to the evaluation". I have written
something about this several times here, one of these posts has been printed in
the latest CSS magazine. I have written this again 2 days ago here on CCC.
The search in a chess program can spoil the knowledge included in his evaluation
function. By doing too much selectivity based on material only (like null move
does for example), a chess program can miss a subtle move. It is not that it
would not have understood the move, it is simply that it had no opportunity to
evaluate it because the pruning algorithm has decided to throw it away.
Open your mind and think differently: a slow program can be very good at
tactics, a fast searcher can be good in positional understanding.
Christophe
>>I think the best way to check knowledge is to look only at games when every
>>program "believe" for many moves that it stands better(it is better to look at
>>slow time control to reduce the affect of tactical mistakes).
>>
>>
>>
>>If a weaker(or at least not stronger) program has better result in these games
>>then it is a clear evidence that it has better positional understanding.
>>
>>The opposite is not truth because the stronger program can have more than 50%
>>because of tactical mistakes of the weaker program.
>>
>>I do not know about results of this test so I have no idea which program has
>>better knowledge.
>>
>> what style of play does it have?
>>>is tiger better with humans too? is tiger's best at blitz or slow chess? how
>>>does it do at defending compared to the best (shredder4!?) and end games
>>>(hiarcs732!?)
>>
>>Why do you think that shredder4 is best at defending and that hiarcs732 is best
>>at endgames?
>>
>>>Is it's nps more like hiarcs or fritz?
>>
>>I read that the number of nps is something like half of the number of fritz.
>>
>>
>>
>>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.