Author: Dann Corbit
Date: 21:34:17 05/09/00
Go up one level in this thread
On May 10, 2000 at 00:23:30, Michael Neish wrote:
>
>Hi,
>
>I'd like to ask what people think of the depth at which programs seem to search
>nowadays. I've read some posts and analyses on here that suggest that some
>commcercial programs habitually search to depth 15 or more in the middlegame,
>whereas other programs seem to take a long time just to get to depth 10 or 11.
>Of course there are differences in search speed, but given the exponential
>nature of the search tree it seems hard for me to believe that these programs
>really are searching so much deeper than their contemporaries.
>
>Are they cheating somehow? Since ply depth is probably a strong selling point,
>as processor speed is for PCs, there is probably a strong motivation to try to
>reach further than your contemporaries, at least on the surface. So, are these
>programmers pruning the trees somehow (apart from the usual techniques talked
>about here), or otherwise eliminating moves so that the search tree doesn't
>expand as quickly? If so, isn't this search depth just illusory? Can
>short-term tactics, well below 15 ply, be missed in this way?
Part of the problem may have to do with the definition of a ply. Different
programs define it differently. But ply depth and strength seem to be very
strongly correlated -- probably the best indicator that there is (at least if we
could all agree to some constant definition for what it means).
Consider two programs which both reach 9 plies searched, but one has quiescent
search, SEE, and various extensions and the other does not. Again, the
definition is misleading. I would like to see a definition like this:
1. minply {due to NULL-move, etc) is the shallowest actual search depth (IOW
brute force depth)
2. aveply {is the actual search depth with assumption that NULL move or
whatever pruning mechanism is sound}
3. maxply {due to extensions, quiescent search, etc} is the maximum depth
actually visited.
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.