Author: Robert Hyatt
Date: 14:57:58 02/08/00
Go up one level in this thread
On February 08, 2000 at 14:24:09, William Bryant wrote: >On February 08, 2000 at 11:41:59, Robert Hyatt wrote: > >>On February 08, 2000 at 10:09:27, William Bryant wrote: >> >>>I need some recomendations on handling time management when the program >>>fails low at the root near the end of the alloted search time. >>>(During game play, not during test suites) >>> >>>How do people handle this? >>> >>>Do you research the PV move at the lower window (which will return a >>>move and a score) and then stop? >> >> >>I re-search... but then pay attention to the score. If the score is more >>than a pawn lower than the last iteration, I use more time, up to 5x more >>time in fact. Hoping that a different ply-1 move will eliminate this loss of >>material. >> > >Thank you for the prompt response. Let me see if I understand. > >You fail low on the root pv move and stop the search, drop the aspiration window >and research the root pv move to get a new score. _If_ this score is >within 1 pawn of the original score, you stop the search and accept this move. >_If_ the score drops more than 1 pawn, a potentially dangerous move, then you >need to try and find a new root pv move and will extend the search up to 5x more >in terms of time to find a new root pv move and to prevent playing a bad move >that will wreck the current position. It isn't done quite that simply. If the score drops by a pawn, I will spend a _bunch_ of time trying to fix it. If the score drops by 1/3 of a pawn, I will spend a significant amount of time trying to fix that, but not nearly as much as I would spend saving material. > >Now, with hopefully good move ordering, do you stop after the first non-pv move >is searched (f the score is good enough on the research), or do you need to >research all of the root moves at this point before determining the new pv move. I stop the search when both of the following is true: (a) I have used the target time _AND_ the score is acceptable; (b) I have used more than the 'panic time allocation' (5x for losing material, 2x for losing positional considerations). > >William >wbryant@ix.netcom.com
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.