Author: Robert Hyatt
Date: 11:45:25 05/04/01
Go up one level in this thread
On May 04, 2001 at 14:06:42, Uri Blass wrote: > >I understand that it is a fail low otherwise you cannot see instantly a clearly >negative score from pondering > >If it is a fail low you know to use more time for the move at depth 17 but >Deeper blue did not use more time to finish the iteration and it was only move >34 so I see no reason not to do it(I can understand not using more time if you >are at move 39 or 40 because it probably cannot help you at 2 hours/40 moves but >not at move 34. You miss the point. At depth=17, _if_ it had had enough time, the search might have failed low. But by the time time ran out, it had _not_ failed low, so it assumed everything was alright and it played the depth=16 move. But it had a _lot_ of information in the hash table from the depth=17 search, and I simply suggest that this information would _definitely_ influence the ponder searches. I will be happy to find a similar example in Crafty's logs from ICC if you want. It doesn't happen every move, or even in every game, but it happens quite often. It is particularly annoying to see a game search produce a score of -1.0 after 14 plies, yet when I set the position up and start from there, it takes 15 or 16 plies to get down to -1. It was all about what was left in the hash table from the previous (incomplete) search. > > ><snipped> >>you make the move and start pondering. >> >>_instantly_ you find the -.5 score as it is left in the transposition table >>from the search that timed out. > >In the case the score was based on d: >-.3 after 0 seconds of pondering at depth 11 >-.66 after 1 second of pondering at depth 13 and >-.5 after 5 seconds of pondering at depth 14 > >It means that the score that was found instantly from pondering was -0.3 > >Here is again the relevant part of the logfile > >hash guess ra1f1Q,Guessing Rxf1 > 7(4) #[h5](-30)[h5](-30) -30v T=0 >Ph6h5 kh1g1 Pb7b5 kg1h1 Kg8f8 kh1h2 Ne5g4 ne3g4N Ph5g4n pf5f6 Rd8d6 kh2g1 Re8e6 >rf1e1 > 7(6) #[h5](-66)############################## -66 T=1 >Ph6h5 kh1g1 Pb7b5 kg1h1 Kg8f8 kh1h2 Ne5g4 ne3g4N Ph5g4n pf5f6 Rd8d6 kh2g1 Re8e6 >rf1e1 > 8(6) #[h5](-50)############################## -50 T=5 > > >Uri
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.