Author: Gerhard Vetter
Date: 15:58:30 03/09/00
Go up one level in this thread
On March 09, 2000 at 12:07:41, James Robertson wrote: >On March 09, 2000 at 09:59:56, Gerhard Vetter wrote: > >>On March 08, 2000 at 19:07:36, James Robertson wrote: >> >>>On March 08, 2000 at 18:34:44, Gerhard Vetter wrote: >>> >>>>On March 08, 2000 at 18:26:44, José de Jesús García Ruvalcaba wrote: >>>> >>>>>On March 08, 2000 at 18:04:48, Andrew Dados wrote: >>>>> >>>>>>On March 08, 2000 at 17:54:35, Gerhard Vetter wrote: >>>>>> >>>>>>>Lately I found this easy-to-solve but elegant little study: >>>>>>> >>>>>>> >>>>>>>8/8/8/8/4k3/1P1p4/1r6/R1N1K3 w Q - >>>>>>> >>>>>>>To my surprise many modern programs can't find the win despite of 5manTBS only >>>>>>>2ply away! >>>>>>> >>>>>>>Junior 6 (with 5manTBS) sees only a draw after 45 min on Celeron 500/ 128 >>>>>>>Also Shredder 4 and Nimzo 7.32 fail to solve it. >>>>>>> >>>>>>>It is interesting that the older versions of some of the above programs need >>>>>>>only 0 sec >>>>>>>(Nimzo99) resp. 10 sec (Junior 5.0). Tiger 12.0e takes also 0 sec. >>>>>>> >>>>>>>It would be interesting to know how other programs do in this position and even >>>>>>>more >>>>>>>interesting if someone can suggest a reason for this odd behaviour of some >>>>>>>leading programs. >>>>>>> >>>>>>>Since I don't know if this information is embedded in the EPD: white has still >>>>>>>castling rights. >>>>>>> >>>>>>>Gerhard >>>>>> >>>>> >>>>> Yes, the queenside castling rights are in the fen string you posted (it is the >>>>>'Q' after the side to move). >>>>> >>>>>>I can be that your tablebases are somehow broken.. note that all programs >>>>>>without TB access solved it... >>>>>>It's of course just a guess. >>>>>> >>>>>>-Andrew- >>>>> >>>>> The very simple explanation is that Nalimov tablebases do not store castling >>>>>rights. >>>>>José. >>>> >>>>sounds like a sensible reason, but why can Fritz 6 solve it then (with >>>>tablebases activated)? >>>>Gerhard >>> >>>Because Fritz only uses the tablebases if the current position is in the >>>tablebases (since it uses tablebases stored on the CD, it would be too slow to >>>use them in the search). >>> >>>James >> >>thanks for everybody's input, what I learned that in some endgame positions one >>better not uses tablebases access because the implementation is still buggy. >>Gerhard > >The only bugs I have found in the Nalimov tablebase implementation have been my >own. :\ > >It was a decision to not store castling rights that Eugene made, so we can't >really call it a bug.... > >James the tablebases themselves work perfect once the 5 piece situation has arrived, I am only wishing that programs like Junior6 etc. would work the transition from their own calculations into the realm of the tbs better. Sorry for trying to call it a bug, as an enduser without programming experience I should rather call it lack of feature. Thanks for all your information! Gerhard
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.