another hard position (mate in 18) from 1998, by Ed Shröder
Posted: Sun Jul 17, 2011 4:31 pm
I'm exhuming another difficult position from CCC's archive. Here's the original message by Ed Shröder:
[d]1r4r1/3q1npk/2b1pbnp/Rp1p4/1N1P3P/2PQ1pP1/1K3B2/5B1R w - - 0 1
And some answers back from the days, showing that no engine could find it 13 years ago, except Gromit, although it didn't see it as a mate :
What would your engine play?
Here's the diagram:Ed Shröder wrote: Subject: Mate in 18 moves?
Author: Ed Schröder
Date: 08:11:14 01/09/98
Mate in 18 moves?
I am busy to release some new EPD test sets in the Rebel9 subscription
area. One the EPD sets is a 200 positions selection from the Austrian
Modul magazine with lots of funny old stuff.
I then came across the following cool "mate in 18(?)" position:
[Event "?"]
[Site "3/93-51 Matt in wieviel?"]
[Date "1992.01.01"]
[Round "?"]
[White "Kristian Skoeld: Sc"]
[Black ""]
[Result "1-0"]
[WhiteElo "?"]
[BlackElo "?"]
[FEN "1r4r1/3q1npk/2b1pbnp/Rp1p4/1N1P3P/2PQ1pP1/1K3B2/5B1R w - - 0 1"]
1. Qxg6+ Kxg6 2. Bd3+ Kh5 3. g4+ Kxg4 4. Rg1+ Kf4 5. Bg3+ Kg4 6. Be1+ Kh3
7. Bf1+ Kh2 8. Bg3+ Kxg1 9. Ra1 Ra8 10. Rc1 Ra1 11. Kxa1 Qa7+ 12. Kb2 Qa1+
13. Kxa1 Ra8+ 14. Na2 Rxa2+ 15. Kxa2 Bg5 16. hxg5 f2 17. Bh3+ f1=Q 18. Rxf1# 1-0
I have tried Rebel9 for a few minutes but it didn't find the mate.
The usual killers (quiet moves) here are:
9.Ra1
10.Rc1
and especially 16.hxg5 f2
Any program?
- Ed -
[d]1r4r1/3q1npk/2b1pbnp/Rp1p4/1N1P3P/2PQ1pP1/1K3B2/5B1R w - - 0 1
And some answers back from the days, showing that no engine could find it 13 years ago, except Gromit, although it didn't see it as a mate :
Covax wrote:Hi,
I test Fritz 5.01, Hiarcs 6.0, Mate 1.0 (mate search engine)
I think, that this test is very hard to today chess programms.
Fritz 5.01 find, if computer x100 faster than Pentium 200 MMX.
Pentium 200 MMX
Fritz 5.01
1.h5 -1.41
(00:06:32)
Hiarcs 6.0
1.h5 -2.14
(00:04:17)
Mate 1.0
Not find in 10 minutes
With best regards, Covax.
I let Prédateur pounder this for a pair of minutes and it sticks to h5. Just like Komodo, Fruit and Stockfish (didn't try other engines on my Mac).Frank Schneider wrote:I tried Gromit and it plays Qxg6 with a score of 0. Of course it can't
see
the mate. The reason for the 0 score is a heuristic that is like 'if
black
has so much more material it should find a variation where the king
isn't
forced to move so often' (anyone else doing things like this? Any
program
playing Qxg6?).
However, if the above position is OK 4. ... Kf4 will be refuted by Bg6
with
mate in 3.
Therefore
4. ... Kh3
5. Bf1+ Kh2
6. Rg4! e5 (or Ra8 7.Nd3 or Qc7 7.Nd3)
7. Bg3+ Kh1/g1
8. Ra1 Ra8
9. Rd1 Ra1
10.Rxa1 Ra8 (or Qa7, Rb1...)
11.Rd1 Ra1
12.Rxa1 Qa7
13.Rb1 Qa1+
14.Kxa1 Bg5
15.Bh3+ Bc1
16.Rxc1#
I didn't analyze deeply, but it still seems to be a mate in 16 that is
easier for humans than for computers.
Frank
What would your engine play?