Author: Sanjiv Karnataki
Date: 06:23:38 02/13/00
Go up one level in this thread
On February 13, 2000 at 08:44:31, Andrew Carey wrote: >On February 13, 2000 at 06:58:08, Enrique Irazoqui wrote: > >>Yesterday I found this 1965 study of Hasek, perfect for a "secret" test set, but >>too beautiful to keep it hiding. I find amazing how the g pawn wins the game and >>also the traveling of black's king to its original e8 square, as if in aeternal >>return. >> >>Of the engines I tried, Nimzo is the fastest in finding 1.b6+ in 44'', followed >>by Hiarcs 8 pre-alpha in 275'', F6a in 432'', and J6a in 554''. Other programs I >>tried didn't solve it in less than 15 minutes on a PIII-500. Enjoy. > >There is no enjoyment. Only for those who believe chess computers are magic and >do not understand them at all. > >To solve this position requires no more than: > >king=infinity >queen=9 >rook=5 >bishop=nite=3 >pawn=1 > >To solve this position relative fast requires no more than extensions on check. >To solve relative faster requires no more than extensions on one-move-only >reply. >It is possible to work out at which depth an unintelligent material only program >will solve it. Please look at Crafty source for these extension types. These >have been for years now. > >b6+ this is a check and does not decrement depth >Kxb6 >Nd7+ this is check and does not decrement depth >Kc6 >Qa6+ this is check and does not decrement depth >KxNd7 this is forced single move reply and does not decrement depth >e6++ this is check and does not decrement depth >Ke8 >Qa8++ this is check and does not decrement depth >Qe8 this is forced single move reply and does not decrement depth >g6 >Qxa8 >g7 >anymove - throwing away the queen will horizon two plys >g8=Q++ > >Eight or nine plies of search is enough. Another two ply for final adjust to >position. Each adjust is with mate threat. Six plies enough if extra extension >on mate in alternate branch and nullmove mate threat detect. The position is >solved with no evaluation function necessary. Computer chess is not magic even >though it may seem like so to some. > >Please do not post positions marked as "great" to a computer chess board when >the position is not interesting. Computer chess is no magic. Computer chess is >quite logical. > >Please do not add to any test set, secret, magic or otherwise. This magic test >set only will only detect for simple programs with no intelligence. This is not >computer chess. > >All is detected by material. Not magic. Not great. > >Thank you. > > Hmm... An interesting lesson in computer chess. As an amateur who started a month ago, I would like to know a few things: a) How many "serious" computer programs use a material only evaluation these days? b) How many "serious" programs evaluate non-captures before captures ? because the best move here is a non-capture. I think that the position is interesting because if it can be solved by just material knowledge, then a more knowledgeable program like "crafty" should figure it out in a snap . But on my Pentium-II/300 crafty has spent 6 minutes with a drawn score reached by the capture move qxc7+ and the subsequent check/onerep/check extensions you so eloquently mentioned. Maybe the interesting thing is to keep computer chess's ability to solve material only positions, as we add more knowledge to it. Of course crafty (a dangerous player by any measure) may still be right and b6 may not be the best move, I will have to see.... > > >> >>Enrique >> >>[D]1NQ5/k1p1p3/7p/pP2P1P1/2P5/2pq4/1n6/6K1 w - - 0 1 >> >>[Event "?"] >>[Site "?"] >>[Date "1965.??.??"] >>[Round "?"] >>[White "Hasek, J."] >>[Black "Study"] >>[Result "1-0"] >>[Annotator ""] >>[SetUp "1"] >>[FEN "1NQ5/k1p1p3/7p/pP2P1P1/2P5/2pq4/1n6/6K1 w - - 0 1"] >>[PlyCount "17"] >> >>1. b6+ (1. Qxc7+ Ka8 $11) (1. Nc6+ Kb6 2. Qb8+ Kc5 3. Qa7+ Kxc4 $19) 1... Kxb6 >>2. Nd7+ Kc6 3. Qa6+ Kxd7 4. e6+ Ke8 5. Qa8+ Qd8 6. g6 Qxa8 7. g7 Qf3 8. g8=Q+ >>Qf8 9. Qg2 1-0
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.