Author: Simon Finn
Date: 13:48:35 06/08/00
Go up one level in this thread
[snip] >>>The critical line is now: >>> >>>1. h6 Kc8 2. h7 Rxh2 3. e7 Kd7 4. b6 Rxh7 5 e8Q+ Kxe8 6. b7 d5 7. b8Q+ Kf7 >>> >>>Black's rook is uncomfortable on h7, but if he has time to play Rh6 >>>the position will be a fortress draw. I don't see how White can >>>prevent this. For example: >>> >>>8. Qc7+ Kf8 9. Qc8+ Kf7 10. Qf5+ Kg8 11. Qxd5+ Kf8 12. Qf5+ Kg8 13. Qxa5 Rh6 >>> >>>and Black just keeps his king on one of the squares f7/f8/g8/h8/h7 >>>while moving his rook between h6 and f6. >>> >>>Simon >> >> Hiarcs may not be the best Analystical program, but Anand who has plenty of >>experience playing against Chess Programs, chose Hiarcs at the Advanced Chess >>Tournament as his analystical software of choice. Has anybody tried to solve it >>with Hiarcs yet? > >I tried a couple of years ago with Hiarcs 6 or 7 on a PII-400 and it didn't find >1.h6 in less than 15 minutes. Since 1. h6 is objectively no better than 1. e7 (both lines lead to forced draws) this could actually be because Hiarcs is very clever (understands the fortress draw after 1. h6 Kc8!) rather than slightly stupid (misses that 1. h6 gxh6? 2. e7 wins for White). However, if I had to guess, I would bet on "stupid". > >Tiger on a PIII-600E with 128MB hash finds h6 after 216 seconds (!), but the >eval is only +0.02. After 30 minutes it's already +0.96, but it doesn't show the >line. In this endgame, a score of +1 isn't particulary significant since Black can hold the draw even with a large material disadvantage. In fact, Black has to sacrifice his d-pawn to acheive the draw. This is due to the fortress position - but evaluating the fortess correctly requires special endgame knowledge which I wouldn't expect any current progam to have. Tablebases help a little, but not enough to recognise that Black can draw from the position after 6. b7. Without 5-man tablebases (and without special endgame knowledge), I would expect most programs to evaluate the final position of my drawing line as about +3, purely on material. (In fact, Crafty without 5-man tablebases evaluates White's advantage at slightly more than 3, so it could resign to a grandmaster from this drawn position. The position is actually still drawn if we give White an extra pawn on h2.) [D]6k1/6p1/7r/Q7/8/8/8/3K4 w - - With 5-man tablebases (but without the 6-man set), I would expect programs to avoid taking the last pawn and produce an evaluation of about +2 instead. However it's still a draw. Or does your program understand this fortress? Simon > >Enrique > >>Pichard
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.