Author: Terry Ripple
Date: 23:49:13 06/24/02
Go up one level in this thread
On June 24, 2002 at 19:05:22, Steve Coladonato wrote:
>On June 24, 2002 at 16:55:03, Terry Ripple wrote:
>
>>On June 24, 2002 at 11:48:46, Steve Coladonato wrote:
>>
>>>On June 24, 2002 at 11:05:29, Robert Hyatt wrote:
>>>
>>>>On June 24, 2002 at 07:35:53, Steve Coladonato wrote:
>>>>
>>>>>There is a post concerning a test position and some analysis by Junior and
>>>>>Shredder. I was curious if there has been a compilation of Depth vs. Time by
>>>>>Engine for some of these test positions where the hardware is the same (or
>>>>>really close to being the same). I'm not so curious as to whether a solution
>>>>>was found, just how long it takes to get to the depth. From the posts, it looks
>>>>>like it took Junior about 1 hour to reach Depth 21 while it took Shredder about
>>>>>1 hour to reach Depth 15. But I don't know if its apples vs. apples with the
>>>>>hardware.
>>>>>
>>>>>Thanks.
>>>>>
>>>>>Steve
>>---------------
>>I believe if program A gets to 21 ply faster than program B it doesn't prove
>>that program A is the better program, because program A could have a very
>>selective search, but with this very selective search it could miss possibly
>>some good moves in a game and cause it to lose sometimes!
>>
>>Regards,
>> Terry
>
>Hi Terry,
>
>In that case, then would not the use of a selective search be considered an
>inferior algorithm to use. I don't know how each program evaluates its score
>for a given position and I do know that the evaluation can be tweaked via
>configuration files. So if the configuration files were ruled out such that
>each program scored the position the same, would not the faster (to a given ply
>depth) program win?
>
>Steve
--------------
Hi Steve,
I would have to say "yes" in response to your question when you word your
question like this!
Though i believe a selective search is most necessary for a program to be able
to search to a certain depth which is necessary in order to beat a certain
strength opponent. But i also feel there is a point where you could make your
program too selective and then you take a chance in missing a possible good move
somewhere in the game which could be crucial! So, by this method of searching
deeper and or faster than the opponent don't necessarly mean better!
Look how well Rebel Century performs which i believe is considered a slower
searcher because of all the knowlege programed into it, which slows the search
down but is still able to perform well even though it is not as fast as some
other top programs.
Regards,
Terry
>>>>Junior doesn't count "plies" the same as everybody else...
>>>
>>>Thanks Dr. Hyatt. There was a pretty big discrepancy there and I was wondering.
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.