Author: Andrew Williams
Date: 07:41:55 08/02/99
Go up one level in this thread
On August 01, 1999 at 21:02:54, Scott Gasch wrote: >Ouch... > >I only can search to 4 ply in that position in 60 sec and my null move succeeds >about 40% of the time. I believe this bad performance has sometime to do with >my qeval routine and the fact that it considers all captures/promotes. I think >I need to be more selective about captures in the qeval routine. My program >plays f2f3 after 1 minute of thought. Here's the report (if anyone cares): > > total nodes.....................................4588297 > time (sec)......................................60.71 > rate (nodes/sec)................................75579 > Your program is nearly twice as fast as mine! But I think you're right about selectivity in qsearch. This is a lot of nodes for 4 ply. You might look carefully at your move ordering as well. > null move successes.............................73 > null move failures..............................117 > null move success rate..........................38.42% > Are these the actual figures? If so, I think there is a problem. Why do you only try null move 190 times out of 4.5 million nodes? In my program it tries null move so long two conditions are satisfied: (a) both sides must have at least two Pieces and (b) the last move wasn't a null move. In this case, it tries null move in about 70% of non-quiesce nodes. If this figure (190) is correct, I think you could be using null move a lot more. > pawn hash hits..................................4469147 > pawn hash misses................................117413 > pawn hash success rate..........................97.44% There's clearly nothing wrong with your pawn hash table! :-) I get a 92% hit rate here. Regards, Andrew
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.