Author: Robert Hyatt
Date: 12:16:57 03/16/01
Go up one level in this thread
On March 16, 2001 at 11:04:01, Miguel A. Ballicora wrote: >On March 16, 2001 at 08:51:52, Robert Hyatt wrote: > >>On March 16, 2001 at 00:23:21, Pham Minh Tri wrote: >> >>>Hi, >>> >>>Suppose the time for searching a move is limited in T minus. And a program has >>>just finished the iterative loop of depth N. My question: How to decide that the >>>time left is enough for the next loop (depth N+1) or not? >>> >>>(Some books suggested me to continue searching next depth and interrupt when >>>time is over. But I don't like this method because of wasting time.) >>> >>>Any suggestions would be very appreciated. >>>Pham >> >> >>Just start the next iteration period. If you don't have time to get a move >>back, you still have time to fail low, which can signal you to use more time >>to find a better move. When you run out of time and make the move, the >>"ponder search" will pick up where you stopped due to everything being saved >>in the hash table... > >This is an idea that I haven't had time to try (of course, most probably it >was tried before by somebody else...). >Suppose this scenario: I have alloted 30 s for the move, I finish one iteration >at 20 s so most probably I won't be able to finish the next one on time. >Would it be useful to go to the next iteration searching with alpha, alpha+1 >just to test if the pv_move fails low? If it doesn't, I just return that move. > >Regards, >Miguel I just search with the normal aspiration window. Fail lows are 30 times easier to compute than normal or fail high moves. Because one move at ply 2 is enough to cause the root to fail low.
This page took 0.01 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.