Author: Robert Hyatt
Date: 17:23:24 03/04/00
Go up one level in this thread
On March 04, 2000 at 15:17:10, Djordje Vidanovic wrote: >On March 04, 2000 at 13:52:43, Ed Schröder wrote: > >>Not having tried this position on my own baby consider this: why add >>extra chess knowledge if these kind of things can so easily seen by >>search. No need to do things twice as the extra knowledge is only >>lowering the NPS of a program. I am not saying this is true for this >>position only that programmers often let do search the work instead >>of adding extra chess knowledge that is not needed anymore now that >>PC's are are so fast and becoming faster and faster and faster..... >> >>As you see in some cases you can even make your program stronger by >>removing chess knowledge. These cases are exceptional but they do >>exist. >> >>Ed > > >I have just got an email inquiring about the performance of Deep Junior on this >position. The sender was quite incredulous so I am attaching DJ's analysis >(reaching up to 900,000 nodes per second). Close to three minutes of analysis >yielded zilch -- so, Ed, fast searchers do not necessarily manage to find key >moves merely through searching deep. > >New position >[D]k7/1b6/8/8/2Q5/8/6p1/6K1 w - - 0 1 > >Analysis by Deep Junior 6.0: > >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qa3+ Kb8 6.Qg3+ Ka8 > +- (5.28) depth: 6 00:00:00 11kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qg7 Kb6 > +- (5.28) depth: 9 00:00:00 112kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qc5+ Kb8 6.Qc3 Ka8 7.Qa3+ Kb8 > +- (5.28) depth: 12 00:00:00 508kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qc5+ Kb8 6.Qc3 Ka8 7.Qa3+ Kb8 > +- (5.28) depth: 15 00:00:03 2864kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qc5+ Kb8 6.Qc3 Ka8 7.Qa3+ Kb8 > +- (5.28) depth: 18 00:00:14 11090kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qc5+ Kb8 6.Qc3 Ka8 7.Qa3+ Kb8 > +- (5.28) depth: 20 00:00:28 22710kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 5.Qc5+ Kb8 6.Qc3 Ka8 7.Qa5+ Kb8 > +- (5.28) depth: 22 00:01:06 52426kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 > +- (5.28) depth: 23 00:01:56 91519kN >1.Qa4+ Kb8 2.Qf4+ Ka7 3.Qf2+ Ka8 4.Qf8+ Ka7 > +- (5.28) depth: 24 00:02:52 134614kN > >(Vidanovic, Naissus 04.03.2000) Two issues: (1) it is only necessary that the program win the game, not win it in the shortest number of moves. You entered a raw position, which means there is no 'context' of what might be repetitions. It will take many checks before it learns which are leading to repetitions, and then starts to migrate the king in the right direction, only to avoid the draws. IMHO this is perfectly ok, just so it wins and doesn't draw. (2) tablebases solve this instantly.
This page took 0.01 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.