What kind of knowledge required for WAC-100 ?

[Vladimir Medvedev]

Geschrieben von: / Posted by: Vladimir Medvedev at 11 February 2004 12:39:44:

When my engine used material-only evaluation, it was capable to solve WAC-100 at ply 20 (nullmove is switched off by late endgame trigger). When all evaluation terms are on, GreKo can't solve the problem at all...
There may be two possible reasons for this:
a) wrong endgame knowledge ( overevaluation of own pawns? )
b) lack of some some significant knowledge
I think the second is more probable, but have no idea which endgame evaluation terms can help here? (GreKo has some knowledge about passed pawns, rule of square, king-passer distance etc... All this does not work...)
Diagram

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 11 February 2004 13:43:27:
Als Antwort auf: / In reply to: What kind of knowledge required for WAC-100 ? geschrieben von: / posted by: Vladimir Medvedev at 11 February 2004 12:39:44:

When my engine used material-only evaluation, it was capable to solve WAC-100 at ply 20 (nullmove is switched off by late endgame trigger). When all evaluation terms are on, GreKo can't solve the problem at all...
There may be two possible reasons for this:
a) wrong endgame knowledge ( overevaluation of own pawns? )
b) lack of some some significant knowledge
I think the second is more probable, but have no idea which endgame evaluation terms can help here? (GreKo has some knowledge about passed pawns, rule of square, king-passer distance etc... All this does not work...)

8/k1b5/P4p2/1Pp2p1p/K1P2P1P/8/3B4/8 w - -
Even at ply 25, Fritz8 prefers Be3 (+5.07) instead of b6+ (+4.78).
Is Be3 wrong ? BigLion likes Be3 from ply 7 up.
/Matthias


BigLion + Taktix

[Lyapko George]

Geschrieben von: / Posted by: Lyapko George at 11 February 2004 13:54:26:
Als Antwort auf: / In reply to: Re: What kind of knowledge required for WAC-100 ? geschrieben von: / posted by: Matthias Gemuh at 11 February 2004 13:43:27:

When my engine used material-only evaluation, it was capable to solve WAC-100 at ply 20 (nullmove is switched off by late endgame trigger). When all evaluation terms are on, GreKo can't solve the problem at all...
There may be two possible reasons for this:
a) wrong endgame knowledge ( overevaluation of own pawns? )
b) lack of some some significant knowledge
I think the second is more probable, but have no idea which endgame evaluation terms can help here? (GreKo has some knowledge about passed pawns, rule of square, king-passer distance etc... All this does not work...)

8/k1b5/P4p2/1Pp2p1p/K1P2P1P/8/3B4/8 w - -
Even at ply 25, Fritz8 prefers Be3 (+5.07) instead of b6+ (+4.78).
Is Be3 wrong ? BigLion likes Be3 from ply 7 up.
/Matthias

Be3 is an alternative solution for WAC-100 which leads to identical position after 1.Be3 Bd6 2.Ka5 Bc7+ 3.b6+ Bxb6 4.Kb5. The key is to find b6 sacrifice anyway...
Best regards,
George

[Slobodan R. Stojanovic]

Geschrieben von: / Posted by: Slobodan R. Stojanovic at 11 February 2004 14:20:58:
Als Antwort auf: / In reply to: What kind of knowledge required for WAC-100 ? geschrieben von: / posted by: Vladimir Medvedev at 11 February 2004 12:39:44:

The next black´s move will surely be Bb6, blocking the position, cutting white´s king penetration and drawing the game.
On the other hand, white´s a6 passed pawn, being on a white square, could be taken only by black king.
So, if white try to preserve its pawn advantage, the game will be draw. Let´s see whether we could sacrifice one pawn and win it.
In this case, after some moves, we´ll have the black king on a6 and the white one on c5, which means: the white king will reach black pawns on the opposite board side first, and win the game.
SL.

[Slobodan R. Stojanovic]

Geschrieben von: / Posted by: Slobodan R. Stojanovic at 11 February 2004 15:55:32:
Als Antwort auf: / In reply to: About the position WAC-100 geschrieben von: / posted by: Slobodan R. Stojanovic at 11 February 2004 14:20:58:

The next black´s move will surely be Bb6, blocking the position, cutting white´s king penetration and drawing the game.

This part of my speculation was wrong. Human has good ideas but sometimes they are disconnected from reality.
SL.

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 11 February 2004 19:44:02:
Als Antwort auf: / In reply to: Re: What kind of knowledge required for WAC-100 ? geschrieben von: / posted by: Lyapko George at 11 February 2004 13:54:26:

When my engine used material-only evaluation, it was capable to solve WAC-100 at ply 20 (nullmove is switched off by late endgame trigger). When all evaluation terms are on, GreKo can't solve the problem at all...
There may be two possible reasons for this:
a) wrong endgame knowledge ( overevaluation of own pawns? )
b) lack of some some significant knowledge
I think the second is more probable, but have no idea which endgame evaluation terms can help here? (GreKo has some knowledge about passed pawns, rule of square, king-passer distance etc... All this does not work...)

8/k1b5/P4p2/1Pp2p1p/K1P2P1P/8/3B4/8 w - -
Even at ply 25, Fritz8 prefers Be3 (+5.07) instead of b6+ (+4.78).
Is Be3 wrong ? BigLion likes Be3 from ply 7 up.
/Matthias

Be3 is an alternative solution for WAC-100 which leads to identical position after 1.Be3 Bd6 2.Ka5 Bc7+ 3.b6+ Bxb6 4.Kb5. The key is to find b6 sacrifice anyway...
Best regards,
George

This is the version from WACNEW.EPD which you can find on my FTP site.
8/k1b5/P4p2/1Pp2p1p/K1P2P1P/8/3B4/8 w - - bm Be3 b6+; id "WAC.100";
There are a number of WAC positions which were busted.


my ftp site {remove http:// unless you like error messages}