Endgame Fireworks II

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Jouni
Posts: 3313
Joined: Wed Mar 08, 2006 8:15 pm

Endgame Fireworks II

Post by Jouni »

[D] 8/8/2b1N3/5p2/3B4/2K5/6kp/R7 w - - 0 1

This is a study, but about as difficult as Shirov thing. Only solution
surprise surprise by :) by Houdini 1.03a x64 2_CPU:
.
.
.
.
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc4 8.Kf2 Be2 9.Ne6 Bg4 10.Ng7 Bh5 11.Bd4 Be8 12.Nf5
² (0.56) Depth: 28/55 00:15:13 3082mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc4 8.Kf2 Be2 9.Ne6 Bg4 10.Ng7 Bh5 11.Bd4 Be8 12.Nf5
² (0.68) Depth: 28/55 00:15:13 3085mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc8 8.Kf2 Bg4 9.Be3 Bf3 10.Ne6 Be4 11.Bd4 Bc6 12.Ng7 Be8 13.Nf5
± (0.96) Depth: 28/55 00:15:14 3089mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc8 8.Kf2 Bg4 9.Be3 Bf3 10.Ne6 Be4 11.Bd4 Bc6 12.Ng7 Be8 13.Nf5
+- (1.90) Depth: 28/55 00:15:17 3100mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bb5 8.Kf2 Be2 9.Ne6 Bg4 10.Ng7 Bh5 11.Nxh5
+- (4.56) Depth: 28/55 00:15:55 3231mN

Jouni
Top
User avatar
michiguel
Posts: 6401
Joined: Thu Mar 09, 2006 8:30 pm
Location: Chicago, Illinois, USA

Re: Endgame Fireworks II

Post by michiguel »

Jouni wrote:[D] 8/8/2b1N3/5p2/3B4/2K5/6kp/R7 w - - 0 1

This is a study, but about as difficult as Shirov thing. Only solution
surprise surprise by :) by Houdini 1.03a x64 2_CPU:
.
.
.
.
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc4 8.Kf2 Be2 9.Ne6 Bg4 10.Ng7 Bh5 11.Bd4 Be8 12.Nf5
² (0.56) Depth: 28/55 00:15:13 3082mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc4 8.Kf2 Be2 9.Ne6 Bg4 10.Ng7 Bh5 11.Bd4 Be8 12.Nf5
² (0.68) Depth: 28/55 00:15:13 3085mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc8 8.Kf2 Bg4 9.Be3 Bf3 10.Ne6 Be4 11.Bd4 Bc6 12.Ng7 Be8 13.Nf5
± (0.96) Depth: 28/55 00:15:14 3089mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bc8 8.Kf2 Bg4 9.Be3 Bf3 10.Ne6 Be4 11.Bd4 Bc6 12.Ng7 Be8 13.Nf5
+- (1.90) Depth: 28/55 00:15:17 3100mN
1.Rh1 Kxh1 2.Kd3 f4 3.Nxf4 Bf3 4.Kc3 Bg4 5.Bc5 Bc8 6.Kd2 Ba6 7.Ke1 Bb5 8.Kf2 Be2 9.Ne6 Bg4 10.Ng7 Bh5 11.Nxh5
+- (4.56) Depth: 28/55 00:15:55 3231mN

Jouni
The PV is seriously wrong. Rh1 Kxh1/ Kd3 and Kg2 wins. The rest is even worst. Rather than finding a solution, it finds a humongous bug that just happen to give Rxh1 as a first move.

Miguel
Top
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: Endgame Fireworks II

Post by Houdini »

Miguel,
michiguel wrote:The PV is seriously wrong. Rh1 Kxh1/ Kd3 and Kg2 wins. The rest is even worst. Rather than finding a solution, it finds a humongous bug that just happen to give Rxh1 as a first move.
The engine is still in "fail-high" state so obviously the PV and score cannot and will not be accurate.
The main line and evaluation will be correct if and when the user lets the engine compute its analysis to the end.

Very nice endgame study, by the way.

Robert
Top
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Endgame Fireworks II

Post by bob »

Houdini wrote:Miguel,
michiguel wrote:The PV is seriously wrong. Rh1 Kxh1/ Kd3 and Kg2 wins. The rest is even worst. Rather than finding a solution, it finds a humongous bug that just happen to give Rxh1 as a first move.
The engine is still in "fail-high" state so obviously the PV and score cannot and will not be accurate.
The main line and evaluation will be correct if and when the user lets the engine compute its analysis to the end.

Very nice endgame study, by the way.

Robert
Then why display a PV that is seriously wrong? Just Rh1! is good enough to get the idea across without providing garbage moves... With a fail high there _is_ no PV. Pulling one out of your hat is not so good..
Top
User avatar
Houdini
Posts: 1471
Joined: Tue Mar 16, 2010 12:00 am

Re: Endgame Fireworks II

Post by Houdini »

The main line for the study at depth 28 shown by Houdini (after 40 minutes using 2 cores) goes as follows:

Code: Select all

28/69	39:39,273	12.730.838.148	+5,65	1.Rh1 Kh3 2.Nf4+ Kg3 3.Bf2+ Kxf4 4.Rxh2 Ke4 5.Rh6 Ba8 6.Rh4+ Kf3 7.Bd4 Bd5 8.Bf6 Kg3 9.Rb4 Be4 10.Kd2 Kf4 11.Ke2 Kg4 12.Be5 Kg5 13.Ke3 Bc6 14.Bf4+ Kg4 15.Rb6
It takes about 5 times longer to produce the correct evaluation and PV than to find that Rh1! is winning, so some patience is needed :).
Top
peter
Posts: 3187
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Endgame Fireworks II

Post by peter »

bob wrote: Then why display a PV that is seriously wrong? Just Rh1! is good enough to get the idea across without providing garbage moves...
Hi Bob!
You mean like this?

8/8/2b1N3/5p2/3B4/2K5/6kp/R7 w - - 0 1

Analysis by IvanHoe T0.4 beta3:

1.Kd3 h1D 2.Txh1 Kxh1 3.Sf4 Kh2 4.Lf2
+/= (0.63) Tiefe: 6 00:00:00 8kN
1.Sf4+ Kg3 2.Sh5+ Kg4 3.Sf6+ Kh3 4.Kd3 h1D 5.Txh1+ Lxh1
+/- (0.84) Tiefe: 6 00:00:00 11kN
...
1.Ta2+ Kh3 2.Sf4+ Kg3 3.Txh2 Kxh2 4.Lf2 La8 5.Kd4 Lc6 6.Kc5 Le4 7.Kc4 Lb7 8.Kd3 La8 9.Kc3 Lb7 10.Lh4 Le4 11.Kd4 Kg1 12.Ke3 Lc6 13.Le1 Kf1 14.Lc3 Le8 15.Le5 Lc6 16.Lf6 Ld7 17.La1
+/- (0.80) Tiefe: 29 00:15:56 3607mN
1.Th1
+/- (0.88 !) Tiefe: 29 00:22:03 5029mN
1.Th1
+/- (0.96 !) Tiefe: 29 00:22:11 5057mN
1.Th1
+/- (1.08 !) Tiefe: 29 00:22:22 5098mN
1.Th1
+/- (1.26 !) Tiefe: 29 00:22:43 5181mN
1.Th1
+- (1.53 !) Tiefe: 29 00:23:03 5255mN
1.Th1
+- (1.93 !) Tiefe: 29 00:23:13 5292mN
1.Th1
+- (2.53 !) Tiefe: 29 00:23:16 5302mN
1.Th1
+- (3.43 !) Tiefe: 29 00:23:20 5315mN
1.Th1
+- (4.78 !) Tiefe: 29 00:23:42 5409mN

:)
4x2,5GHz, 1024Mb hash, no bbs
Peter.
Top
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Endgame Fireworks II

Post by bob »

peter wrote:
bob wrote: Then why display a PV that is seriously wrong? Just Rh1! is good enough to get the idea across without providing garbage moves...
Hi Bob!
You mean like this?

8/8/2b1N3/5p2/3B4/2K5/6kp/R7 w - - 0 1

Analysis by IvanHoe T0.4 beta3:

1.Kd3 h1D 2.Txh1 Kxh1 3.Sf4 Kh2 4.Lf2
+/= (0.63) Tiefe: 6 00:00:00 8kN
1.Sf4+ Kg3 2.Sh5+ Kg4 3.Sf6+ Kh3 4.Kd3 h1D 5.Txh1+ Lxh1
+/- (0.84) Tiefe: 6 00:00:00 11kN
...
1.Ta2+ Kh3 2.Sf4+ Kg3 3.Txh2 Kxh2 4.Lf2 La8 5.Kd4 Lc6 6.Kc5 Le4 7.Kc4 Lb7 8.Kd3 La8 9.Kc3 Lb7 10.Lh4 Le4 11.Kd4 Kg1 12.Ke3 Lc6 13.Le1 Kf1 14.Lc3 Le8 15.Le5 Lc6 16.Lf6 Ld7 17.La1
+/- (0.80) Tiefe: 29 00:15:56 3607mN
1.Th1
+/- (0.88 !) Tiefe: 29 00:22:03 5029mN
1.Th1
+/- (0.96 !) Tiefe: 29 00:22:11 5057mN
1.Th1
+/- (1.08 !) Tiefe: 29 00:22:22 5098mN
1.Th1
+/- (1.26 !) Tiefe: 29 00:22:43 5181mN
1.Th1
+- (1.53 !) Tiefe: 29 00:23:03 5255mN
1.Th1
+- (1.93 !) Tiefe: 29 00:23:13 5292mN
1.Th1
+- (2.53 !) Tiefe: 29 00:23:16 5302mN
1.Th1
+- (3.43 !) Tiefe: 29 00:23:20 5315mN
1.Th1
+- (4.78 !) Tiefe: 29 00:23:42 5409mN

:)
4x2,5GHz, 1024Mb hash, no bbs
Looks reasonable to me. If you get a fail high at the root, there is no concept of a "best move" at ply=2. Which means there is no way to figure out which is the best move at ply=3, and so forth...
Top
peter
Posts: 3187
Joined: Sat Feb 16, 2008 7:38 am
Full name: Peter Martan

Re: Endgame Fireworks II

Post by peter »

bob wrote: Looks reasonable to me. If you get a fail high at the root, there is no concept of a "best move" at ply=2. Which means there is no way to figure out which is the best move at ply=3, and so forth...
Problem with PV seems to be 1... Kxh1, which might be a mate in 12

8/8/2b1N3/5p2/3B4/2K5/6kp/R7 w - - 0 1

1. Rh1 Kxh1 2. Nf4 Bf3 3. Kd2 Bg2 4. Ke2 Be4 5. Kf2 Bc2 6. Be3 Bd1 7. Ne6 Bc2
8. Ng7 f4 9. Bxf4 Bg6 10. Be5 Bf5 11. Nh5 Bc2 12. Ng3#

but the declining of the sac is some longer:

1. Rh1 Be4 2. Kd2 Bb7 3. Ke2 Kh3 4. Nf4+ Kg3 5. Bf2+ Kxf4 6. Rxh2#25

or

1. Rh1 Kh3 2. Nf4+ Kg3 3. Bf2+ Kxf4 4. Rxh2#24
Peter.
Top

Return to “Computer Chess Club: General Topics”