Author: Richard Pijl
Date: 02:47:54 05/13/05
Go up one level in this thread
On May 13, 2005 at 05:15:07, Alvaro Jose Povoa Cardoso wrote: >Hi, >I saw that you removed the code that adjusts the bounds in Hashprobe(). >Could you please elaborate on that? Was it wrong or unsafe before? >You had it for a long time, it sure must be sound. > >Best regards, >Alvaro Cardoso I had this code some time ago as well, and the main problem I found with it is that it could make it harder to fight instability in search. Suppose you have found a lowerbound of alpha+10 (and beta still higher than that), it is possible that the search will fail low though, eg with (in the case of crafty) alpha+10. What value would you return? alpha+10? In the parent node this won't be recognized as a fail-high, but as a new pv score. So the few nodes that are saved by using this mechanism does cause a little problem. Whether this is still useful to use is probably dependent on the program. I removed it. Richard.
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.