Author: Miguel A. Ballicora
Date: 10:06:46 08/09/01
Go up one level in this thread
On August 09, 2001 at 12:36:35, TEERAPONG TOVIRAT wrote: >> >> I don't use the TT to detect repetitions. Instead, I check in the move list, >>looking for the hash key of the actual position on the board. >> And I don't store draw-by-repetition scores in the TT, because they are messy >>to handle. > > >Hi, > >Accidentally, this post lead to the same question as mines below. >This is in my mind. I plan to avoid storing draw by repetition scores >in hashtable. Did you run some tests about this ? How about the result? >Please,let me know. I can tell you what I do: I do not store in the hashtable positions whne a repetition was "just" detected. In other words, I have a flag when I call search(.. ,.. ,.. ,&rep_flag). When it detects a repetition, at the beginning of search, it returns draw and sets the rep_flag=1. So, I know that this move caused a repetition in the following node. If this move that causes the draw is the best move, I do not store the draw, I reset the rep_flag = 0 and I return draw. So, if I ever visit this node again, it will be very easy to see if it is really a draw. One ply search would do it, so lacking the hashtable hit won't do much harm. If this node is not a draw because it comes from another path, the absence of a hash hit is a blessing. This does not solve all the problems but I think it solve many of them without hurting the efficiency of the hashtable a lot. When I tested it, it helped to improve the performance in the FINE70 problem. I have a compiler switch to turn this off. If anybody is interested, I can do some experiments with some positions that you think could be critical. Regards, Miguel > >Thanks, >Teerapong
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.