Author: Dave Gomboc
Date: 16:00:28 03/07/99
Go up one level in this thread
On March 07, 1999 at 16:34:47, Robert Hyatt wrote: >On March 07, 1999 at 04:50:13, Dave Gomboc wrote: > >>On March 02, 1999 at 17:25:08, Will Singleton wrote: >> >>> 5R2/8/8/4rp2/3r4/5k2/2B1p2P/2K1R3 w - - >>> White to move and draw. >>> >>>Well, that one turned out to be more interesting than I anticipated. We found >>>that a couple of programs could see the draw w/o tablebases (Ferret, and perhaps >>>Lambchop), and Crafty gave the tablebase draw result. Gives the rest of us >>>something to think about. >>> >>>Aside from that, there was pretty much agreement on Rxf5. One program chose >>>Bxf5 with a larger negative score than the concensus on Rxf5. That might bear >>>looking into. >>> >>>Zchess was the only one not to consider Rg8. The passed pawns were scored so >>>highly that the idea of bringing the rooks down to defend wasn't even an option. >>> >>>And for some reason, it took my program a ton of nodes to resolve the fail-low, >>>and then a ton more to find the better move. Something for me to worry about. >>>(Maybe time to put in some endgame knowledge, but I think there's something else >>>going on.) >>> >>>Will >> >>How are you trying to resolve the fail-low at the PV? Have you tried completely >>restarting the search (with the TT intact! :) My understanding is that this is >>the preferred method nowadays... but certainly, YMMV. >> >>I am interested to hear from those who do restart on PV fail-lows, and those who >>tried it but found it lacking. >> >>Dave Gomboc > > >I have done this two ways: > >(1) first root move fails low, continue searching, hoping that another move >will not fail low and you get something useful. If you search all moves and >the score == alpha, you have to relax alpha and search again. > >(2) first root move fails low, relax alpha immediately and search it again to >get the true score for the first move. > >I do (2) at present, for a couple of reasons: (a) I want to know how far the >score dropped so I have some idea of how much trouble I am in and how much extra >time I might want to allocate; (b) It is not uncommon to 'win' a pawn in early >iterations only to see later that you can't hold on to it. So the old best >move is still best, but with a much lower score. It is more efficient to re- >search it immediately than to try all the other moves (which will also fail >low) and then try again. > >Which is better is a good question. I have seen cases where I wished I would >just continue searching, and others where I am glad I re-search after the first >move fails low. I don't just mean option (2). I suggest restarting the iterative deepening, at ply ONE! Shocking, I know. :-) But the TT will carry most of the information you need to get back up quickly. It will guide the search to investigate alternatives if they don't suck, or still cut them off if they're even worse than the PV that failed low. Dave Gomboc
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.