Author: Uwe Immel
Date: 11:38:59 03/09/98
Go up one level in this thread
On March 09, 1998 at 05:57:27, Kai Lübke wrote: >On March 08, 1998 at 05:39:33, Uwe Immel wrote: > >>On March 06, 1998 at 07:31:30, Kai Lübke wrote: >> >>>The following position (probably not unknown to many who followed >>>the WMCCC in Paris) is from the Dark Thought vs. MChess 7 game in Paris. >>>In this situation, MChess 7 played Rc4!! >>> >>>2r2kn1/1p3pr1/pq2p3/3pP2N/b4P1P/PnPQ4/1B4P1/1R2KB1R b K - ; bm Rc4; id >>>"shep06" > >>By trying the mentioned position, I noticed some kind of strange >>behavior >>of Junior. >> >>In lack of Fritz having an infinite level, I normally use the fixed >>depth level (=99) for long analysis. This dosn't work proper by using >>the >>Junior engine, because it takes a few minutes 'til Junior shows any >>information. >>For example setting ply depth 15, I have to wait about 14 minutes until >>Junior shows his evaluation and pv, at depth 14 it almost takes 4 >>minutes. The span of time increases with every further ply, so I didn't >>try ply 99 ;-) >> >>Not a huge problem so far as you can alternative use the time per >>move level, but then the following difference comes up, that at fixed >>ply depth 13 Junior finds the move Rc4 in 2:55 (Cyrix 166+) and at >>level 3600 sec/move the keymove is not found within 17 ply ! >>At level 600 sec/move Rc4 is found at ply 14 in 7:58. >>Either Junior does another search on fixed depth level or it's a kind >>of coincidence by forcing Junior to think only to ply 13. >> >>However, I'm not sure which level to use to get the (suppposed) best >>move in shortest time (especially for longer analysis). > >I use "Infinite Analysis" (Alt-F2) for my testing and Junior solved the >above position at 4:50 in ply 15. >It's possible that Junior readjusts some internal values (read: the way >it searches) depending on the given reflection time - Amir is probably >the only authority to comment on this. Seems so..., at level "game in 60" Junior finds the Rc4 also at ply 13. >BTW, any reason why using "Infinite Analysis" for longer analysis is >bad? No, it's more an old habit from earlier times when you had a "real" infinite level, why I use the fixed ply depth level 99 (that's somehow the same way), permanent analysis fits the same. >I think I heard something about a 999 second limit or something? If I remember correctly, 99999 secs is the max time for the correspondence analysis, max time per move should be 99999 secs, but Fritz adjusts this to 22463 secs as well as other higher time settings, that's a flaw, but not a serious bug for me ;-) regards Uwe
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.