Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Is the NPS for minimax devided by NPS in alpha-beta = 5 ?

Author: leonid

Date: 19:30:25 06/16/00

Go up one level in this thread


On June 16, 2000 at 16:51:09, Dann Corbit wrote:

>On June 16, 2000 at 15:36:09, leonid wrote:
>[snip]
>>Very interesting! I just found above your numbers and they are very close.
>>Exactly like you say. My number is 400% different. Strange!
>
>Sounds like a bug.  I'd run it through a profiler and see where all the time is
>going.

Please, say later what you found.

The most likely is that efficency of generation of the nodes (legal moves) for
each ply goes down dramatically when only few nodes in each ply are used. This
is why this strange number 5 is more that real. In the same time, if this number
is significantly lower that 5, it could be indicative of very inefficient move
ordering. Efficiency of the program is inverse to the efficency of its move
generation. This is why this "silly number" is so good to be watched closely.

Leonid.





This page took 0.04 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.