Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How can moveordering efficiency be measured?

Author: Michel Langeveld

Date: 03:49:09 12/31/01

Go up one level in this thread


On December 31, 2001 at 05:37:33, Matthias Gemuh wrote:

>On December 31, 2001 at 05:26:11, Michel Langeveld wrote:
>
>>I want to check how efficient Nullmover is.
>
>
>Divide "fail_highs_on_first_move" by "total_fail_highs".
>I score about 88% ordering on BWTC test suite. This is not good enough.
>Others seem to get over 95%. Tell me what you get.
>
>Regards,
>Matthias.

I think the way you and Dann postes to measure fail highes is the same isn't it?

   for (int i=0; i< ml.number; i++)
   {
      totalMoves++;
      doWhiteMove(ml.moves[i]);
      val = -qsearchBlack(-beta, -alpha);
      restoreWhiteMove(ml.moves[i]);

      if (val >= beta)
      {
         //count the scores
         betaCounter++;
         if ( i == 0) firstMoveWasBest++;
         betaCounterNth += i+1;

         return beta;
      }
      if (val > alpha)
      {
         alpha = val;
      }
   }

I get 93% in a complete 10 ply search in this position, that took me 56.51
seconds.

[D]rnbqkbnr/ppp2ppp/8/3pp3/3PP3/8/PPP2PPP/RNBQKBNR w KQkq -

I get 95% in a complete 11 ply search in the same position.
Which takes 288.88 seconds.

Michel



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.