Page 2 of 4

Re: Yet Another Mate Solving Test

Posted: Tue Jan 18, 2011 1:43 pm
by Jouni
You need better mate solvers :)

Analysis by Critter 0.90 64-bit:

1.Na3 Rhe8 2.Ne5+ Qxe5 3.Bxe5 Rxe7 4.Qxe7 Rc8 5.Rd1 Ba8 6.Rd7 Re8 7.Rd6+ Kb7 8.Rb6+ Ka7 9.Qxc7+ Bb7 10.Qxb7#
+- (#10) Depth: 27/54 00:17:40 3988mN

Analysis by Houdini 1.5a x64:

1.Na3 Rhe8 2.Ne5+ Qxe5 3.Rxc7+ Kb6 4.Nc4+ Ka7 5.Nxe5 Ka8 6.Qc5 Rxe5 7.Bxe5 Rd8 8.Qb6 Rd1+ 9.Rxd1 h6 10.Qxb7#
+- (#10) Depth: 26/61 00:06:35 2175mN

I suggest someone with spare CPU time tests all 8+ movers with these 2 engines and enough time to find errors!

Jouni

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 2:49 am
by Dann Corbit
Does anyone have the full solution to the following:
2nB1k2/2p3pp/3p2p1/n2K4/3Qp1N1/1Np4R/rb1p4/1Rq3r1 w - -
2nB1k2/2p3pp/3p2p1/n2K4/3Qp1N1/1Np4R/1b1p4/1Rq3r1 w - -
8/1p6/2B1K1p1/6NP/8/8/1P4pq/1R4bk w - -
8/k1B4K/3N4/pppppppp/8/8/1b6/5R2 w - -
R7/1Kpkr1q1/2p1pBP1/2p2pP1/3pr1p1/QP2pb2/1n4Nb/1B6 w - -

I have this as a mate in 13:
r4b2/3pp2K/r1pp2p1/B2k2NP/2nP2Q1/2P5/P1Pp4/3N3b w - -
Does anyone have a mate in 11 for it?

For all of these, I have them as mate in 10 rather than mate in 11:
5N2/4p3/8/4R3/6P1/2P1Rp1p/PKP1nppP/3k1bn1 w - -
6k1/3N4/6K1/7n/8/B7/8/8 w - -
k4b2/Pnp1p1p1/pB2P1Pp/P2B4/1p6/1P5K/7p/8 w - -
K7/8/4p3/3p4/8/5N1B/8/6Bk w - -
1R4K1/8/8/8/p7/8/p7/k4B1N w - -

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 7:32 am
by Michael Diosi
Did you try the retrograde option of Chest on your "solution" ?



Michael
http://www.playwitharena.com

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 8:03 am
by Michael Diosi
Hi,



I only tried this one:
FEN: 6k1/3N4/6K1/7n/8/B7/8/8 w - - 0 1

ChestUCI:
Parameter-Analyse: Ausgangs-Stellung ! W:3/17
11 00:00 0 0 +M11 1.Ba3-c1 Nh5-g7 2.Nd7-f6+ Kg8-h8 3.Bc1-h6 Ng7-f5 4.Bh6-f8 Nf5-e7+ 5.Kg6-f7 Ne7-f5 6.Nf6-e4 Kh8-h7 7.Ne4-g5+ Kh7-h8 8.Bf8-a3 Nf5-d6+ 9.Kf7-f8 Nd6-b5 10.Ba3-b2+ Nb5-d4 11.Bb2xd4+


and it is a mate in 11 not in 10.

Michael
http://www.playwitharena.com

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 3:07 pm
by Ignacio
Dann Corbit wrote:Does anyone have the full solution to the following:
2nB1k2/2p3pp/3p2p1/n2K4/3Qp1N1/1Np4R/rb1p4/1Rq3r1 w - -
2nB1k2/2p3pp/3p2p1/n2K4/3Qp1N1/1Np4R/1b1p4/1Rq3r1 w - -
8/1p6/2B1K1p1/6NP/8/8/1P4pq/1R4bk w - -
8/k1B4K/3N4/pppppppp/8/8/1b6/5R2 w - -
R7/1Kpkr1q1/2p1pBP1/2p2pP1/3pr1p1/QP2pb2/1n4Nb/1B6 w - -

I have this as a mate in 13:
r4b2/3pp2K/r1pp2p1/B2k2NP/2nP2Q1/2P5/P1Pp4/3N3b w - -
Does anyone have a mate in 11 for it?

For all of these, I have them as mate in 10 rather than mate in 11:
5N2/4p3/8/4R3/6P1/2P1Rp1p/PKP1nppP/3k1bn1 w - -
6k1/3N4/6K1/7n/8/B7/8/8 w - -
k4b2/Pnp1p1p1/pB2P1Pp/P2B4/1p6/1P5K/7p/8 w - -
K7/8/4p3/3p4/8/5N1B/8/6Bk w - -
1R4K1/8/8/8/p7/8/p7/k4B1N w - -
Hi Dann,
Do you have the last Chest database?
Almost all positions (the 3rd NO) are there with parameters to find solutions.

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 4:04 pm
by Jouni
I have sent some correction to Franz and have feeling, that 2-5% of Chest database positions are incorrect e.g. there is a shorter mate. One idea: add MATERIAL only mode to Houdini/Critter/etc for faster mate finding. There is one in Goliath engine.

Jouni

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 4:22 pm
by F.Huber
Dann Corbit wrote: Does anyone have the full solution to the following:
2nB1k2/2p3pp/3p2p1/n2K4/3Qp1N1/1Np4R/rb1p4/1Rq3r1 w - -
2nB1k2/2p3pp/3p2p1/n2K4/3Qp1N1/1Np4R/1b1p4/1Rq3r1 w - -
8/1p6/2B1K1p1/6NP/8/8/1P4pq/1R4bk w - - ???
8/k1B4K/3N4/pppppppp/8/8/1b6/5R2 w - -
R7/1Kpkr1q1/2p1pBP1/2p2pP1/3pr1p1/QP2pb2/1n4Nb/1B6 w - -
With one exception all these positions are in the ChestUCI database,
so you can solve each of them within a few seconds.
The exception is
8/1p6/2B1K1p1/6NP/8/8/1P4pq/1R4bk w - -
and I'm really wondering which program sees a mate here? :shock:
I have this as a mate in 13:
r4b2/3pp2K/r1pp2p1/B2k2NP/2nP2Q1/2P5/P1Pp4/3N3b w - -
Does anyone have a mate in 11 for it?
Yes, the ChestUCI database! :wink:
For all of these, I have them as mate in 10 rather than mate in 11:
5N2/4p3/8/4R3/6P1/2P1Rp1p/PKP1nppP/3k1bn1 w - - #10
6k1/3N4/6K1/7n/8/B7/8/8 w - - bm #10; #11, NO #10
k4b2/Pnp1p1p1/pB2P1Pp/P2B4/1p6/1P5K/7p/8 w - - #10
K7/8/4p3/3p4/8/5N1B/8/6Bk w - - bm #10; #11, NO #10
1R4K1/8/8/8/p7/8/p7/k4B1N w - - bm #10; #11, NO #10
All these problems are also in the ChestUCI database, 3 of them are
only #11 (it can be proved even in brute force within a few minutes
that they are not #10!)

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 4:43 pm
by F.Huber
Jouni wrote:I have sent some correction to Franz and have feeling, that 2-5% of Chest database positions are incorrect e.g. there is a shorter mate.
2-5% ? That would be 100-300 mates??? :shock:
Well, I'm sure that's much too high!

Of course there are some incorrect mates, but it's not ChestUCI's fault but the fault of the problem composers!
If I find a mate problem and the composer states it is a mate in N, then I try to find a mate in N, and if ChestUCI finds such a solution, then I usually see no reason trying to find a mate in N-1, N-2, etc. ...
That would take just too much time on my quite slow computer.

Just today I've updated the ChestUCI database on my website, where I've fixed all incorrect mates which were reported in the last time.

But I don't trust mate announcements of normal chess engines very much, e.g. Rybka reported three #9 in Jouni's test suite in this thread, which were in fact #11 (checked even in brute force mode).
So if someone finds any shorter mate than reported in the ChestUCI database with a normal chess engine, then he should try to get a solution with ChestUCI and post this solution with the special parameters and solution time found, otherwise I won't be able to prove this here on my system.

Regards,
Franz

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 7:28 pm
by Paloma
[D]1r5r/1bp1Rppp/p1kB4/1p1q4/1Q6/2P2N2/PP3PPP/RN4K1 w - - 0 1
1.Na3 Rhe8 2.Ne5+ Qxe5 3.Bxe5 Rxe7 4.Qxe7 Rc8 5.Rd1 Ba8 6.Rd7 Re8 7.Rd6+ Kb7 8.Rb6+ Ka7 9.Qxc7+ Bb7 10.Qxb7#

Code: Select all

Analysis by Critter 0.90 64-bit:

1.Na3 Rhe8 2.Ne5+ Qxe5 3.Bxe5 Rxe7 4.Qxe7 Rc8 5.Rd1 Ba8 6.Rd7 Re8 7.Rd6+ Kb7 8.Rb6+ Ka7 9.Qxc7+ Bb7 10.Qxb7#
+- (#10) Depth: 27/54 00:17:40 3988mN
After 1.Na3 Rhe8 Chest5.1UCI find mate in 9

[D]1r2r3/1bp1Rppp/p1kB4/1p1q4/1Q6/N1P2N2/PP3PPP/R5K1 w - -
Engine: ChestUCI Ver.5.1 C0/R0/K4/P9/X38
von Franz Huber

9.02 0:30 +M9 2.Se5+ 706

9/09 2:36 +M9 2.Se5+ Dxe5 3.Txc7+ Kb6 4.Sc4+ Ka7
5.Sxe5 Txe5 6.Lxe5 Tc8 7.Dc5+ Ka8
8.Txc8+ Lxc8 9.Dxc8+ Ka7 10.Ld4+ 722

Bester Zug: Sf3-e5 Zeit: 2:37.171 min K/s: 722.698 Knoten: 113.327.792

so it would be possible the start position is mate in 10.

The P parameter must now be 10, but it takes unfortunately a long time.
.

Re: Yet Another Mate Solving Test

Posted: Fri Jan 21, 2011 7:35 pm
by F.Huber
Paloma wrote: The P parameter must now be 10, but it takes unfortunately a long time.
Well, testing this position I've already reached parameters K3/P10/X44.

Is your K4 really necessary? Then maybe K4/P10/X44 could work ...