Author: Christophe Theron
Date: 09:37:37 10/17/98
Go up one level in this thread
On October 17, 1998 at 05:23:04, blass uri wrote:
>
>On October 17, 1998 at 04:57:00, Alessio Iacovoni wrote:
>
>>Gambisoft posts:
>>
>>"Customers report unisono that JUNIOR 5 runs at top speed. We can
>>confirm this. With 350 MHz the search depth starts double digit most
>>of the time!"
>>
>>Results of the engine seem to support the claim that it is a very strong engine.
>>However, when running it against Fritz 5 and especially Hiarcs on Junior
>>platform I've got the impression that, fast as it may be, it is not intelligent
>>enough... sometimes it seems to make moves that not even a patzer would make.
>
>There are positions that every program does sometime moves that not even a
>patzer would make.
>
>I assume that you are talking about positions that Junior5 does mistakes and
>other programs do not do mistakes.
>
>Can you post some positions that you are talking about?
>
>I assume that you are talking about the early stage of the opening.
>It usually does not cause Junior problems against computers because usually
>Junior is in book in these positions.
>
>>
>>It may go 2 or 3 plies ahead of Crafty or Fritz.. but at the detriment of
>>position and basic chess rules.
>
>The depth of Junior5 is not the number of plies.
>
>The depth is something close to the number of captures+the number of threat
>moves+2*the number of regular moves.
>
>I do not think that it is tactically stronger than fritz5(The reason that it has
>good results against computers is mainly positional reason).
>
>Uri
The search depth is not a good indication of a program's strength.
For example I know several small changes (generally 1 or 2 lines of code
changes) that give Tiger one more ply in depth, or even 2 plies more.
Very impressive when you look at the main lines on the screen. You have the
impression that this version is going to smoke any other program.
But when you try to prove it by playing a lot of real games, you just realize
that the program does not play better. Or even worse, that you have lost some
strength in the process.
So every programmer has to make choices in this field. The only thing I'm
looking for is strength. The resulting average search depth is just a "side
effect" of these optimizations, and sometimes a better version is a version that
searches less deep. This has been the case when I switched from Tiger 11.3 to
Tiger 11.4.
Amir has a strong program that seems to search very deep. The same strength can
also be achieved by a program with a shallower search. It's just a question of
tuning.
In the computer area, and especially in the computer chess area, you can achieve
the same goal by different paths. Take a look at CSTal for example.
Christophe
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.