Author: Ed Schröder
Date: 23:37:24 01/21/01
Go up one level in this thread
On January 22, 2001 at 01:58:34, Howard Exner wrote: >On January 22, 2001 at 00:45:09, Howard Exner wrote: > >>On January 21, 2001 at 19:59:54, James T. Walker wrote: >> >>>I get almost the identical thing on my Athlon 900Mhz. I measured it to be >>>almost 2.5x faster than a K6-2-450. My machine always finds d6 at 1:25 and >>>plays 12....d6 at 1:48 (set for 40/1hour). Perhaps it was accidentally forced >>>to move early? In any case it is strange that Century would even consider such >>>a move. But it's not the only strange move I've seen Century make. >>>Jim >> >>To confirm your findings I followed the idea of manually playing Deep Fritz's >>moves. Then let Rebel think and enter its own choices at 40/1:00 tournament >>mode, like you on an Athlon 900. For Fritz I entered the moves immediately until >>it went out of book on move 12. Here the log file had Deep Fritz moving >>Nf3 after 193 seconds so I entered the move at 90 seconds to compensate for the >>machine differences. Sure enough Rebel played not Rg8 but d6. It had O-O way >>down on the 18th iteration. >> >>To more accurately discover what the auto/manual inconsistency is, a 40/2 manual >>input of Fritz's moves, using identical times as the pgn, on a 450 AMD through >>DOS is required. DOS because that is what SSDF is using for Rebel. > >Tried the same experiment on my K6-2 500 under windows. >Mimicing Fritz's moves and times it took to play these moves >on move 12 Rebel plays d6. The original pgn file states that Rebel plays Rg8, >using a total of 381 seconds. Manually on my machine it plays d6 using a total >of 476 seconds. > >Another entry that looks suspect is Rebel's move 10. Not so much the move chosen >of Nc6, but the time used to play this first move out of Rebel's book ... >281 seconds. That seems too short a time for Rebel out of book as typically it >takes a good amount of time to enter an initial out of book move. On the K6-2 >500 for example it actual plays d6 on move 10 using 476 seconds. This supports >the theory that some auto232 oddity is interupting Rebel's thought process and >making it move before it is ready. As Jeff pointed out also. > >The parameter that makes Rebel move quicker, as in the last three Van der Wiel >games, is not used in SSDF is it? The algorithm (playing fast) is excluded from auto232 games as it was developed to play humans. Ed >In any case duplicating how it plays with auto232 on Tony's 450 versus manual >is difficult since I do not have the exact machine and settings that he has.
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.