Author: Peter McKenzie
Date: 20:54:52 09/07/01
Go up one level in this thread
On September 07, 2001 at 22:15:32, Jim Monaghan wrote: >Hi Peter, > >Here is a difficult R+P endgame for programs illustrating the power of connected >passers vs scattered ones. > >[D] 8/5k2/R5pp/3pP2P/p2r1PK1/8/8/8 b - - 0 > >Karpov - Deep Thought, 1990. > >Deep Thought played the greedy 1...gxh5+?? and Karpov quickly showed the power >of connected passed pawns in this type of ending: > >1... gxh5+ 2. Kf5! (Not 2. Kxh5? Rxf4) Kg7 3. Ra7+ Kf8 4. e6 Re4 5. Rd7 Rc4 6. >Rxd5 h4 7. Rd3 >Ke7 8. Rd7+ Kf8 9. Rh7 h5 10. Ke5 h3 11. f5 Kg8 12. Rxh5 a3 13. Rxh3 a2 14. Ra3 >Rc5+ 15. Kf6 (1-0) > >Absolutely essential was 1...g5! to break up White's pawns at all costs, viz. > >1... g5 2. Rxh6 Rxf4+ (Not 2...gxf4? 3. Kf5!) 3. Kxg5 Rf1 should hold. > >(Source: Practical Endgame Play, N. McDonald) > >I tried this on a few engines and they seem very confused. I suppose it would >take a very long think on an engine's part to work it all out, if at all. >Rather than solving it thematically (intuiatively?) as a human would. Knowledge >vs Brute Force. It's possible that positions like this would show more of an >engine's shortcomings as opposed to tactical tricks in most endgame studies. Yes, its the type of position that is a good test of positional knowledge. But I'm not 100% convinced that gxh5 is so bad. After 1...gh 2.Kf4 how about 2...h4!? I analysed this with LambChop for a while, and it is by no means clear. Black even wins if white is not careful. While I wouldn't be surprised if white has a win, I would be equally unsurprised if black is holding. GMs (and McDonald isn't the strongest GM around) can easily make mistakes in these complex endgames. Peter > >Cheers, >Jim
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.