Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Score of moves becomes identically as depth increases ?

Author: Jose Santos

Date: 14:20:52 09/09/03

Go up one level in this thread


Hi again,

>So, are you saying that a pure minmax search delivers the same values ?

Yes, that's what I'm saying and I'm pretty sure about it (unfortunately).

>How about the node counts in either case ?

What do you mean by this ? how many nodes I search for each move ?
Here is it at ply 4 (first is the move, second is the score, third is the number
of boards analyzed to determine it, without alpha-beta cuts).

G1-F3  -220  9754
G1-H3  -220  8883
B1-A3  -220  8885
B1-C3  -184  9757
H2-H3  -184  8457
H2-H4  -184  9329
G2-G3  -184  9347
G2-G4  -184  9328
F2-F3  -184  8457
F2-F4  -184  8929
E2-E3  -184 13280
E2-E4     0 13308
D2-D3     0 11961
D2-D4     0 12437
C2-C3     0  9347
C2-C4     0  9818
B2-B3     0  9347
B2-B4     0  9332
A2-A3     0  8457
A2-A4     0  9329

>That's hard to believe for me. IMHO, grouping the scores like in your case
>above, looks to me very much like a consequence of alpha-beta.

Yes, I know. I also couldn't believe on it, that's why I made this post.

>If in your case, alpha-beta and minmax-pure deliver the same set of scores on
>the root level, then something is definitely going wrong.

Yes... but it doesn't play stupidly bad (altought it's not very good either). If
you are curious you can download latest version here:
http://ctp.di.fct.unl.pt/~jcas/chess/chesswin32.exe
I've also done a Linux version:
http://ctp.di.fct.unl.pt/~jcas/chess/chesslinux.exe
Win32 version is more recent. Both are just console for now, I've done a GUI
yet.

Regards,

                                                 Jose Santos



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.