Author: Christophe Theron
Date: 10:31:59 05/19/98
Go up one level in this thread
On May 18, 1998 at 18:58:49, Komputer Korner wrote:
>Perhaps Amir Ban would carefully explain the depth numbers of Junior's
>search. Even though I consider Junior to have the most accurate score of
>a position, it's search depth numbers do not make sense unless they are
>selected lines of a selective searcher in which case they are not as
>useful to the user. The user needs information on full width depth in
>plies.
Amir said a number of times that Junior uses half plies depth, but
Junior is very selective and ply depth doesn't make sense any more.
Would you ask a grandmaster to announce the ply depth he is currently
doing? :)
All we can do is take several programs, make them play games against
each other at fixed depth, and try to figure out at what "depth" they
seems to play at equal strength.
It still seems to make sense because most top programs are using the
same kind of search (trying to catch most things inside a certain
horizon, then to catch only hanging pieces outside of this horizon), but
it could become obsolete in the future if somebody manages to come
closer to the way humans think. CSTal is already hard to categorize.
Christophe
>On May 16, 1998 at 02:12:56, Christophe Theron wrote:
>
>>On May 15, 1998 at 03:07:56, Moritz Berger wrote:
>>
>>>On May 14, 1998 at 18:23:08, Christophe Theron wrote:
>>>
>>>>>The depth of Junior always amazes me, just like genius depth (although
>>>>>genius does only print out its positional depth, it seems to see 2 ply
>>>>>more).
>>>>
>>>>I agree with that. Genius searching to N plies is equivalent (from a
>>>>strength point of view) to a brute force search on N+2 or N+3 plies. Say
>>>>N+2.5 in average.
>>>>
>>>>
>>>> Christophe
>>>
>>>Maybe Genius is doing some static evalutations, e.g. static exchanges to
>>>make it appear that way?
>>>
>>>Moritz
>>
>>No, it is really searching deeper than the depth it displays. You can
>>see it in the computed lines.
>>
>>
>> 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.