Author: Gerrit Reubold
Date: 08:51:01 08/03/00
Go up one level in this thread
On August 03, 2000 at 10:51:31, Mogens Larsen wrote: >On August 03, 2000 at 10:25:29, Gerrit Reubold wrote: > >>Hi Mogens, > >Hello Gerrit, > >>About the importance of storing results from previous searches: >>Bringer doesn't do this (instead the hashtables are always cleared between the >>moves). :-) >> >>I implemented it this way to be able to always reproduce searches (and bugs). > >Okay, thank you for the information. I did have the suspicion that I was talking >rubbish :o). > >>If Bringer's game was played with winboard, you might want to examine the >>Bringer.pgn file which is created in Bringer's directory. This file stores all >>winboard games, including information about Bringer's score's, time to move... > >Here's the Bringer pgn from the game in question. There's a twenty second >discrepancy, but the gap might have several explanation. However, I won't >venture a guess. I'm surprised by the extent of the time usage. What made >Bringer use that much time? Bringer did play 8 moves from the book. That leaves 40 min. for 32 moves, about 75 sec/per move on average. Bringer (I think most engines) will use/need more time on fail high or fail low, I think 105 sec. is not so surprising. BTW: these 105 sec. are Bringer's time to play Bxh6, not the time to change the PV to Bxh6. About the 20 sec. discrepancy: Are you using the same hash table size? Do you analyze in 1-best mode? Greetings, Gerrit
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.