Author: Robert Hyatt
Date: 19:54:34 08/13/01
Go up one level in this thread
On August 13, 2001 at 22:20:18, Artem Pyatakov wrote: >Well, After tweaking my time control and some of my history settings, it seems >like I have pondering under control and it DOES help my engine go about one ply >deeper when it guesses the right move - thanks to people who responded to my >last post about pondering. > >I was wondering about a scenario, however, and about how people handled it in >their engine (just for fun, I will lay this out like a traditional math/physics >problem): > >Initial conditions: >Let's say the normal search produced a PV of length=1, because the very first >move produced a hit in the hash (transposition table). > >Problem: >Right now in my engine, I do not ponder in these cases, because I have no idea >what opponent move to ponder about (it's not in the PV). > I try two things: (1) probe the hash table. If you get a hit, and it has a suggested best move, ponder that. It isn't always good, but it is better than nothing. (2) flip sides, do a short search, and use the move that returns as the move to ponder. Crafty calls this a "puzzling search" and you will see the message "puzzling for a move to ponder" on rare occasions. >Possible Solutions: >1) I am guessing that it is often possible to discover the "next" move in the PV >by simply looking at the transposition table once again and seeing if we can >extend the PV further > >2) May be I should try to ponder about a different move??? I just don't like the >idea of my engine wasting time... > >3) Any other ideas about how this problem can be addressed? I know this is >probably not a very important thing, but I was just curious about a possible >solution. > >Thank you. > >Artem
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.