Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A new look.

Author: Dan Andersson

Date: 03:05:38 06/20/03

Go up one level in this thread


On June 20, 2003 at 05:41:04, Dan Andersson wrote:

> I tried to look at it in a new way. Reflect upon this:
> d(t) is the depth reached by the perfectly ordered tree were t is nodes or
>time.
> dm(t) is the depth of the minimax tree.
> dm(t)=2*d(t)
> di(t,k) is a tree in the middle.
> di(t,k)=(1+k)*d(t) k is in [0, 1]
> di(t,k1)<di(t,k2) iff k1<k2
  di(t,k1)<di(t,k2) iff k2<k1
> what properties does di have?
> The gain is linear to search depth. A doubling in search depth gives a doubling
>in gained depth. 2*di(t,k2)-2*di(t,k1)=2(1+k2)*d(t)-2*(1+k1)*d(t)=2*(k2-k1)*d(t)
> The gain is constant to time. A doubling in time gives a constant addition in
>gained depth. (k2-k1)*(d(2*t)-d(t))=(k2-k1)*constant since d is a logarithic
>function of t.
> This might also be horribly wrong. Feel free to comment.
Very tired right now :)
Only problem with this analysis is that k is not defined properly. But if move
ordering is close to perfect this won't matter for reasonable search depths.
Then k will be close to zero anyway. Maybe I'll lok into that after some sleep.
>
>MvH Dan Andersson



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.