Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hashtable problems

Author: Daniel Clausen

Date: 07:38:11 09/11/02

Go up one level in this thread


On September 11, 2002 at 08:09:52, Steffen Basting wrote:

>Hello!
>I've got some problems with my hashtable...
>Here's the output of the starting position:

[snip]

>705603 nodes, 101 seconds: 71497 nodes/sec
>q-nodes: 6587101        evals: 6587101

Since #nodes is higher than #qnodes, I assume that with nodes you mean
non-qnodes. Still, having almost 10x more qnodes than other non-qnodes from the
initial position is a bit strange, to say the least. ;) [I'd say about 10-20% in
the initial and up to 50% in tactical positions sounds reasonable - to me]


>5       12      569305          687     g5g6  -
>5       12      2047277         689     a1b2 a3b2 b2a3 a3a2 a2a3 b2a1 a1a2
>                                        a2b2 b2a2 b3b2 b1b2 b2b1 b1b2   +
>16      12      2898223         695     g5g6 h7g6 f5f6 g7f6 h5h6 f6f5 h6h7

>Something is wrong - I simply don't find the mistake...Have a look at the line
>of depth 12 for example - if the search had stopped at this point, my program
>would had lost in a won position with a positive score.

Even worse, the move a1b2 is illegal. (not to mention the rest of the PV :)


>Any suggestions - what might be the mistake?

When retrieving a bestmove from the HT, you do test whether it's really a legal
move in this position? Since two distinct positions can have the same hashkey
(not just the part for the slotnumber in the HT, the whole 64 bit), you always
have to test the bestmove in the HT for legality in the current position.

That or I misread your search-log somehow.

Sargon



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.