Test your engine - mate in 7, problem 189 by M.R. Vukcevic

[Slobodan R. Stojanovic]

Geschrieben von: / Posted by: Slobodan R. Stojanovic at 15 February 2004 14:58:52:

7R/n1p1pBk1/2r2pP1/1qN5/3rP1PP/4p1Kb/1Bn5/bQ6 w - - 0 1

[Peter Fendrich]

Geschrieben von: / Posted by: Peter Fendrich at 15 February 2004 17:22:23:
Als Antwort auf: / In reply to: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Slobodan R. Stojanovic at 15 February 2004 14:58:52:

7R/n1p1pBk1/2r2pP1/1qN5/3rP1PP/4p1Kb/1Bn5/bQ6 w - - 0 1

This is Terra on my laptop.
It seems that both Rh7 and Rg8 is winning but Rg8 gives mate.
Even if I force Rg8 Terra can't find the wonderful Qf1!! with mate after Rg8 Kh6, in a decent time.
/Peter
Terra UCI 52 MB:
3 00:01 0,00 1.Rg8+ Kh6 2.Rh8+ Kg7
4 00:01 0,00 1.Rg8+ Kh6 2.Rh8+ Kg7
5 00:01 0,00 1.Rg8+ Kh6 2.Rh8+ Kg7
6 00:01 0,00 1.Rg8+ Kh6 2.Rh8+ Kg7
7 00:01 0,00 1.Rg8+ Kh6 2.Rh8+ Kg7
8 00:01 0,00 1.Rg8+ Kh6 2.Rh8+ Kg7
8 00:05 0,61 1.Rh7+
8 00:05 1,52 1.Rh7+ Kf8 2.Qa2 e6 3.Bxe6 Bxb2 4.g7+ Ke7 5.g8Q+ Kd6 6.Kxh3 Rxc5 7.Qxa7 Rxe4
9 00:07 1,52 1.Rh7+
9 00:12 1,97 1.Rh7+
9 00:13 3,10 1.Rh7+ Kf8 2.Bxd4 Bxg4 3.Bb3 e6 4.Nd7+ Ke8 5.Nxf6+ Kd8 6.g7 Qxb3 7.Qxb3
10 00:16 3,10 1.Rh7+
10 00:19 3,55 1.Rh7+
10 00:43 5,49 1.Rh7+ Kf8 2.Bxd4 Bxg4 3.Bb3 Re6 4.Nxe6+ Bxe6 5.g7+ Kg8 6.Bxe6+ Kxh7 7.g8Q+ Kh6 8.Bxe3+ Nxe3 9.Qxa1

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 15 February 2004 17:43:30:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Peter Fendrich at 15 February 2004 17:22:23:

7R/n1p1pBk1/2r2pP1/1qN5/3rP1PP/4p1Kb/1Bn5/bQ6 w - - 0 1

This is Terra on my laptop.
It seems that both Rh7 and Rg8 is winning but Rg8 gives mate.
Even if I force Rg8 Terra can't find the wonderful Qf1!! with mate after Rg8 Kh6, in a decent time.

Gothmog (on a Pentium IV 2.4 GHz) also has difficulties here. It plays Rh7+ instantly. After
47 seconds, it gives a score of mate in 9 for Rh7+. Finally, after more than two minutes,
it switches to Qf1 with mate in 7:

1 0.00 0.00 138 Ne6+ Rxe6 Rg8+
2 0.00 0.03 2787 Ne6+ Rxe6 Rg8+ Kh6 Rh8+ Kg7
3 3.03 0.11 10345 Rh7+ Kf8 Nd7+ Rxd7 Rh8+ Kg7 Rg8+ Kh6 g5+
Qxg5+ hxg5+ Kxg5 Kxh3 Bxb2 Qxb2
4 3.72 0.50 66034 Rh7+ Kf8 Nd7+ Rxd7 Rh8+ Kg7 Rg8+ Kh6 g5+
Qxg5+ hxg5+ Kxg5 Kxh3 Bxb2 Qg1+ Kf4 Qg4+
5 9.94 5.26 777355 Rh7+ Kf8 Qa2 e6 Bxe6 Rd7 Nxd7+ Ke8 g7 Qe5+
Nxe5
6 23.44 12.19 1854908 Rh7+ Kf8 Bxd4 e5 Bd5
7 >= #40 41.90 6288550 Rh7+
7 #9 47.98 7193841 Rh7+ Kf8 Qf1 Bxf1 Rh8+ Kg7 Rg8+ Kh6 g5+
fxg5 Rh8+ Kg7 Rh7+ Kf8 Nd7+ Rxd7 Rh8#
8 #9 54.00 8215520 Rh7+ Kf8 Qf1 Bxf1 Rh8+ Kg7 Rg8+ Kh6 g5+
fxg5 Rh8+ Kg7 Rh7+ Kf8 Nd7+ Rxd7 Rh8#
9 #9 1:04 9969529 Rh7+ Kf8 Qf1 Bxf1 Rh8+ Kg7 Rg8+ Kh6 g5+
fxg5 Rh8+ Kg7 Rh7+ Kf8 Nd7+ Rxd7 Rh8#
10 >= #8 2:03 20167421 Qf1 Kxh8 Qf5 Kg7 Qh5 e6 Qh7+
10 >= #7 2:24 24220075 Qf1 Kxh8 Qf5 Kg7 Qh5 e6 Qh7+
10

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 15 February 2004 18:04:08:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Tord Romstad at 15 February 2004 17:43:30:

7R/n1p1pBk1/2r2pP1/1qN5/3rP1PP/4p1Kb/1Bn5/bQ6 w - - 0 1

This is Terra on my laptop.
It seems that both Rh7 and Rg8 is winning but Rg8 gives mate.
Even if I force Rg8 Terra can't find the wonderful Qf1!! with mate after Rg8 Kh6, in a decent time.

Gothmog (on a Pentium IV 2.4 GHz) also has difficulties here. It plays Rh7+ instantly. After
47 seconds, it gives a score of mate in 9 for Rh7+. Finally, after more than two minutes,
it switches to Qf1 with mate in 7:

Unfortunately movei has no chance because it was not built to solve mates.
After finding mate in 9 by Rh7+ it simply stops to search.
It is not interested in finding shorter mate after finishing the iteration.
When the score is not mate my aspiration window is simply score of previous iteration-30,score of previous iteration+30(It may be better to change it to be not constant but it is not in the top of my priorities)
It is clearly illogical to use this aspiration window in case that the score of previous iteration is mate and I did not consider it as important doing special programming for the case that the score of previous iteration is mate.
Uri

[Peter Fendrich]

Geschrieben von: / Posted by: Peter Fendrich at 15 February 2004 19:59:22:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Tord Romstad at 15 February 2004 17:43:30:

7R/n1p1pBk1/2r2pP1/1qN5/3rP1PP/4p1Kb/1Bn5/bQ6 w - - 0 1

This is Terra on my laptop.
It seems that both Rh7 and Rg8 is winning but Rg8 gives mate.
Even if I force Rg8 Terra can't find the wonderful Qf1!! with mate after Rg8 Kh6, in a decent time.

Gothmog (on a Pentium IV 2.4 GHz) also has difficulties here. It plays Rh7+ instantly. After
47 seconds, it gives a score of mate in 9 for Rh7+. Finally, after more than two minutes,
it switches to Qf1 with mate in 7:

1 0.00 0.00 138 Ne6+ Rxe6 Rg8+
2 0.00 0.03 2787 Ne6+ Rxe6 Rg8+ Kh6 Rh8+ Kg7
3 3.03 0.11 10345 Rh7+ Kf8 Nd7+ Rxd7 Rh8+ Kg7 Rg8+ Kh6 g5+
Qxg5+ hxg5+ Kxg5 Kxh3 Bxb2 Qxb2
4 3.72 0.50 66034 Rh7+ Kf8 Nd7+ Rxd7 Rh8+ Kg7 Rg8+ Kh6 g5+
Qxg5+ hxg5+ Kxg5 Kxh3 Bxb2 Qg1+ Kf4 Qg4+
5 9.94 5.26 777355 Rh7+ Kf8 Qa2 e6 Bxe6 Rd7 Nxd7+ Ke8 g7 Qe5+
Nxe5
6 23.44 12.19 1854908 Rh7+ Kf8 Bxd4 e5 Bd5
7 >= #40 41.90 6288550 Rh7+
7 #9 47.98 7193841 Rh7+ Kf8 Qf1 Bxf1 Rh8+ Kg7 Rg8+ Kh6 g5+
fxg5 Rh8+ Kg7 Rh7+ Kf8 Nd7+ Rxd7 Rh8#
8 #9 54.00 8215520 Rh7+ Kf8 Qf1 Bxf1 Rh8+ Kg7 Rg8+ Kh6 g5+
fxg5 Rh8+ Kg7 Rh7+ Kf8 Nd7+ Rxd7 Rh8#
9 #9 1:04 9969529 Rh7+ Kf8 Qf1 Bxf1 Rh8+ Kg7 Rg8+ Kh6 g5+
fxg5 Rh8+ Kg7 Rh7+ Kf8 Nd7+ Rxd7 Rh8#
10 >= #8 2:03 20167421 Qf1 Kxh8 Qf5 Kg7 Qh5 e6 Qh7+
10 >= #7 2:24 24220075 Qf1 Kxh8 Qf5 Kg7 Qh5 e6 Qh7+
10 10 #7 2:25 24438064 Qf1 Kxh8 Qf5 Kg7 Qh5 e6 Qh7+ Kf8 Bxe6 Ke8

I wouldn't call this difficulties!
Your tactical monster solves it with one of the possible mating lines.
The line I was thinking about was 1.Rg8 Kh6 2.Qf1 ... with a somewhat different mate pattern from Got's and Terra couldn't even find a mate in 2 minutes after 1.Rg8 Kh6.
It's nice to see how you manage to create a tactical machine that performs so well in practical play. I suppose that you will add on more and more positional knowledge which will make Gothmog really strong.
The only games where Terra match Got is when the tactical shots are blocked (or Got sacrifice himself to death) and where the the resulting position is in favour for Terra.
My impression is that Got is already fairly strong in the endgame.
/Peter

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 15 February 2004 22:47:12:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Peter Fendrich at 15 February 2004 19:59:22:

I wouldn't call this difficulties!
Your tactical monster solves it with one of the possible mating lines.
The line I was thinking about was 1.Rg8 Kh6 2.Qf1 ... with a somewhat different mate pattern from Got's and Terra couldn't even find a mate in 2 minutes after 1.Rg8 Kh6.
It's nice to see how you manage to create a tactical machine that performs so well in practical play. I suppose that you will add on more and more positional knowledge which will make Gothmog really strong.
The only games where Terra match Got is when the tactical shots are blocked (or Got sacrifice himself to death) and where the the resulting position is in favour for Terra.
My impression is that Got is already fairly strong in the endgame.

Hi Peter,
Thanks for your kind words. To a certain extent, Gothmog's tactical abilities is
just an illusion. It has huge speculative king safety scores, and often guesses
that a combination works one or two plies before it is able to calculate it all.
Most of the time, the guesses are correct, but it also happens that they are wrong.
This is when "Got sacrifice himself to death".
Of course I hope to add more positional knowledge, but it is not easy. I have reached
the frustrating stage when most of the changes I do to my engine just make it play
worse than before.
By the way, a few days ago, I browsed through (mainly for nostalgic reasons) old
issues of "PLY", the magazine of the SSDF. One of the things that caught my eye
was the chess programming tutorial included in 1989-1990. The tutorial was
in some respects very far ahead of its time, for instance it was advocating bitboards
many years before they became mainstream.
I remembered this tutorial very well, because it was the first technical introduction
to computer chess I ever read. I did not, however, remember the name of the
author. It was therefore a great and pleasant surprise to discover that the tutorial
was written by Peter Fendrich!
Thank you, Peter, for being the person who introduced me to chess programming!

Tord

[Peter Fendrich]

Geschrieben von: / Posted by: Peter Fendrich at 15 February 2004 23:38:38:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Tord Romstad at 15 February 2004 22:47:12:

I wouldn't call this difficulties!
Your tactical monster solves it with one of the possible mating lines.
The line I was thinking about was 1.Rg8 Kh6 2.Qf1 ... with a somewhat different mate pattern from Got's and Terra couldn't even find a mate in 2 minutes after 1.Rg8 Kh6.
It's nice to see how you manage to create a tactical machine that performs so well in practical play. I suppose that you will add on more and more positional knowledge which will make Gothmog really strong.
The only games where Terra match Got is when the tactical shots are blocked (or Got sacrifice himself to death) and where the the resulting position is in favour for Terra.
My impression is that Got is already fairly strong in the endgame.

Hi Peter,
Thanks for your kind words. To a certain extent, Gothmog's tactical abilities is
just an illusion. It has huge speculative king safety scores, and often guesses
that a combination works one or two plies before it is able to calculate it all.
Most of the time, the guesses are correct, but it also happens that they are wrong.
This is when "Got sacrifice himself to death".
Of course I hope to add more positional knowledge, but it is not easy. I have reached
the frustrating stage when most of the changes I do to my engine just make it play
worse than before.
By the way, a few days ago, I browsed through (mainly for nostalgic reasons) old
issues of "PLY", the magazine of the SSDF. One of the things that caught my eye
was the chess programming tutorial included in 1989-1990. The tutorial was
in some respects very far ahead of its time, for instance it was advocating bitboards
many years before they became mainstream.
I remembered this tutorial very well, because it was the first technical introduction
to computer chess I ever read. I did not, however, remember the name of the
author. It was therefore a great and pleasant surprise to discover that the tutorial
was written by Peter Fendrich!
Thank you, Peter, for being the person who introduced me to chess programming!

Tord

No, no please don't bring that article up! I'm not very proud of it and I think I did better things for SSDF...
I used bitboards several years before that on a Texas computer (TI 99 I think) with 16-bits word-length. There was a lot of glueing to acheive 64 bits with all its boolean operations and of course it was a performance killer, but it was fun!
Now when the future is here and the 64-bits processors is soon to be standard I'm fed up with pure bitboards and will try something else
/Peter

[Roger Brown]

Geschrieben von: / Posted by: Roger Brown at 16 February 2004 02:48:34:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Tord Romstad at 15 February 2004 22:47:12:

Hi Peter,
Thanks for your kind words. To a certain extent, Gothmog's tactical abilities is
just an illusion. It has huge speculative king safety scores, and often guesses
that a combination works one or two plies before it is able to calculate it all.
Most of the time, the guesses are correct, but it also happens that they are wrong.
This is when "Got sacrifice himself to death".

Hello Tord,
I am willing to bet that your estimation of your engine's capacities are somewhat on the modest side. I simply must tell you that I intend to have a deeper look at it and expose you for the extremely modest person that you are.

Gothmog has something, different about it.
Later.

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 16 February 2004 13:32:05:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Peter Fendrich at 15 February 2004 23:38:38:

By the way, a few days ago, I browsed through (mainly for nostalgic reasons) old
issues of "PLY", the magazine of the SSDF. One of the things that caught my eye
was the chess programming tutorial included in 1989-1990. The tutorial was
in some respects very far ahead of its time, for instance it was advocating bitboards
many years before they became mainstream.
I remembered this tutorial very well, because it was the first technical introduction
to computer chess I ever read. I did not, however, remember the name of the
author. It was therefore a great and pleasant surprise to discover that the tutorial
was written by Peter Fendrich!
Thank you, Peter, for being the person who introduced me to chess programming!

No, no please don't bring that article up! I'm not very proud of it and I think I did better things for SSDF...
I used bitboards several years before that on a Texas computer (TI 99 I think) with 16-bits word-length. There was a lot of glueing to acheive 64 bits with all its boolean operations and of course it was a performance killer, but it was fun!
Now when the future is here and the 64-bits processors is soon to be standard I'm fed up with pure bitboards and will try something else

Considering the time it was written, I don't think it was that bad. There
wasn't a lot of chess programming resources at the time. Your article was
nothing to be ashamed of, and it certainly helped to whet my appetite.
As you know, I am heading in the opposite direction, and it is possible that I
will give (probably non-rotated) bitboards a try later this year.
Tord

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 16 February 2004 13:37:23:
Als Antwort auf: / In reply to: Re: Test your engine - mate in 7, problem 189 by M.R. Vukcevic geschrieben von: / posted by: Roger Brown at 16 February 2004 02:48:34:

Hi Peter,
Thanks for your kind words. To a certain extent, Gothmog's tactical abilities is
just an illusion. It has huge speculative king safety scores, and often guesses
that a combination works one or two plies before it is able to calculate it all.
Most of the time, the guesses are correct, but it also happens that they are wrong.
This is when "Got sacrifice himself to death".

Hello Tord,
I am willing to bet that your estimation of your engine's capacities are somewhat on the modest side. I simply must tell you that I intend to have a deeper look at it and expose you for the extremely modest person that you are.

Gothmog has something, different about it.

Hello!
I'm afraid you will be disappointed, especially if you test it mainly against
The Baron. The Baron is Gothmog's nightmare opponent number one, and tends to
win by crushing margins.
Thanks, this is a great compliment. I want it to be different, more than I
want it to be strong.
Tord