Author: Robert Hyatt
Date: 09:03:48 10/23/97
Go up one level in this thread
On October 23, 1997 at 10:53:29, Chris Whittington wrote: > >On October 23, 1997 at 10:41:18, Robert Hyatt wrote: > >>On October 23, 1997 at 10:20:58, Chris Whittington wrote: >> >> >>> >>>The next iteration it does the same thing. >>>The next iteration the same thing again. >>>The next iteration it is running out of time, does the same thing again, >>>but the time controller decides to stop the search after the fai-high >>>for Ng2+, but before the ensuing fail-low. So, because the search times >>>out during a critical window, CSTal actually decides to play Ng2+ and >>>loses of course :( >>> >>> >>>Merde ! as they say in Paris :) >>> >>>Chris >>> >> >>this is solvable too. When time runs out, I *never* bail out of the >>search until I start to search *another* root move. I allow a 2X time >>overflow here in fact, because I handle the case where the first move is >>good, but the second one is going to be better. I make sure that >>whatever >>root move I am "hung" on when time runs out gets searched to completion >>to >>see if it is better, or if it going to fail low. > >hmmmm. all very well, but their are times when I've given myself panic >time and then more panic time and then this happens; there has to be a >time when you just have to bail out, no ? > >Chris quite possibly. But adding that "qualification" makes the problem much less frequent. But there's always the case of two moves, where A is better than B, but each iteration closes the gap, and the node count for B keeps going up, until the last iteration where B will replace A *if* there is enough time. I do my best to allow this...
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.