Author: Chris Whittington
Date: 04:50:35 11/21/97
Go up one level in this thread
On November 20, 1997 at 23:55:56, Howard Exner wrote: >Here are 5 positions to test on your computer programs. The theme >is how to play when the bishop is the opposite color of the queening >square. > >k7/pp6/8/P7/3K4/P7/P4B2/8 w - - id"a3-a4 - Wrong Bishop" bm a3 a4; >5k2/8/7P/8/8/8/K6P/1B6 w - - id"B-h7 - Wrong Bishop" bm b1h7; >8/7p/6p1/3k4/B5PP/8/8/1K6 w - - id"a4e8 - Wrong Bishop" bm a4e8; >8/6p1/8/3kP2P/6K1/8/8/2b5 w - - id"h5-h6 - Wrong Bishop" bm h5h6; >8/8/pp6/3b2K1/1P6/PN2k3/8/8 w - - id"Na5 draw - Wrong Bishop" bm b3a5; > >The first three are wins and the last two are draws. > >I tested these on Rebel 8. The results are: > >K6-233 60Mb Hash > >1. Not solved by 15:00 >2. 0:51 >3. 0:16 >4. 0:21 >5. Not solved by 15:00 > >Problems 1 and 5 are difficult for Rebel 8. A known issue >with Rebel 8 is it's difficulty with doubled pawns on the a >or h files. An old Maresch test position uncovered this. > >Could others post some of there results for these 5 problems. >Maybe 1 and 5 are difficult for all programs? CSTal version 287_d, P6 200, 16 Mb hash 1. a4 33s 2. Bh7 0s 3. Be8 86s 4. h6 30s 5. unsolvable positions 1-4 are all solvable with the knowledge programmed in. Basically the prg needs to know that the king can get to cover the queening square in time; but there are some tricky sepcial cases to consider. 1-4 are examples where search alone will tale quite soem time, if ever, to reach the solution. its an issue of time spent: a) coding b) slowing up the program by looking for the knowledge c) getting the special cases right (else you'll lose drawn positions, and draw won positions. 5. is semi-unsolvable. Its virtually unsolvable by search, and its also unsolvable by 'normal' knowledge and search. You would need the very special case knowledge that to win the black king must run off and capture the white rook file pawns; this diversion would then allow the white king to cover the queening square. I wouldn't like to have to code this sort of knowledge, not even for $1,000,000. But without such knowledge, and with the 'normal' knowledge; search doesn't really help; search finds the problem fairly quickly, but horizons it away by keeping the black king in the centre (holds off the white king), and wiggles around the bishop - this horizoning can go on for many many plys. Chris Whittington
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.