Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: DB will never play with REBEL, they simple are afraid no to do well

Author: Robert Hyatt

Date: 20:53:31 10/15/99

Go up one level in this thread


On October 15, 1999 at 19:11:48, Ratko V Tomic wrote:

>> IE alpha/beta alone reduces the effective branching factor
>> from 38 to roughly sqrt(38) == 6.  Null move can reduce this by 1/2.
>
>Does Crafty has really an effective branching factor of 3 in typical middle
>games in the full width search? Also, what's your guess on the proportion of
>junk searched in such case. For the sake of simplicity we could define "junk"
>anything that is, say, 1/2 pawn worse than the root node value (it could be
>the running value of PV for simpler test). Would the percentage of the non-junk
>nodes (e.g. the nodes within 1/2 pawn from PV value) go exponentially to zero
>with depth? (That was the original point of disagreement in this small
>exchange.)

yes, 3 is pretty typical, although it steadily drops as material comes off.

as far as 'junk' I have no way of measuring that, since alpha/beta doesn't
provide 'how bad' a move is compared to another, just 'better' or 'worse'.


>
>I think it would be interesting if you could find time to rig your development
>Crafty with few counters and get the proportions of (non-)junk in relation to
>depth and also possibly its dependence on the delta (such as 1/2 pawn)
>separating the junk from the non-junk.

the problem is junk vs non-junk.  I use PVS, which means except for a few dozen
nodes, all the rest are searched with a null window.  So that it is impossible
to determine how much worse a move is than another...



This page took 0.01 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.