That's great, thank you!
In the German forum of CSS Jörg Oster started tests with a modified Stockfish. May be you like to read:
http://forum.computerschach.de/
Thread: Modifizierte Engines für geänderte Pattwertung gesucht
Could you please provide an UCI exe-file to me? Unfortunately I am unexperienced with any kind of programming.
Request for engines with modified stalemate values
Moderators: hgm, Rebel, chrisw
-
- Posts: 45
- Joined: Sat Jun 21, 2014 7:00 pm
-
- Posts: 4833
- Joined: Sun Aug 10, 2008 3:15 pm
- Location: Philippines
Re: Request for engines with modified stalemate values
Thanks for pointing out the link, but I can't understand German language, tried it in google translate and it helps a bit.Steamhammer wrote:That's great, thank you!
In the German forum of CSS Jörg Oster started tests with a modified Stockfish. May be you like to read:
http://forum.computerschach.de/
Thread: Modifizierte Engines für geänderte Pattwertung gesucht
Could you please provide an UCI exe-file to me? Unfortunately I am unexperienced with any kind of programming.
I have provided the download link in my previous post where you can download the exe file.
-
- Posts: 45
- Joined: Sat Jun 21, 2014 7:00 pm
Re: Request for engines with modified stalemate values
Here is an interesting example for how changed stalemate values could apply.
It's an engine game played in the IC engine hall. White has clear advantage, as acknowledged by Black's Stockfish (1.52/46), and it is White's turn to move. White's evals are much higher (3.32/51), as he is running a modified Stockfish with changed stalemate values. There seems to be no way for White to win this game, so the engine favours to stalemate Black by transforming the position into an endgame K+P vs K.
Unfortunately, after reaching the advantageous endgame the "auto draw" functions rules on a draw and prevents White from stalemating Black.
The example shows two things:
1) A typical position, where advantage is not big enough to mate, but to force a stalemate (by zugzwang).
2) The auto draw function should not apply for centaur tours, as far as it is based on "insufficient material" (with the only exception K vs K).
It also shows, that a modified engine is able to evaluate the chances of stalemate, but it is still an open question, whether in other cases it might handicap itself as going for a stalemate, when it would still have chances to achieve a mate. Also it's not clear, how playing without tablebases affects the chances.
It's an engine game played in the IC engine hall. White has clear advantage, as acknowledged by Black's Stockfish (1.52/46), and it is White's turn to move. White's evals are much higher (3.32/51), as he is running a modified Stockfish with changed stalemate values. There seems to be no way for White to win this game, so the engine favours to stalemate Black by transforming the position into an endgame K+P vs K.
Unfortunately, after reaching the advantageous endgame the "auto draw" functions rules on a draw and prevents White from stalemating Black.
The example shows two things:
1) A typical position, where advantage is not big enough to mate, but to force a stalemate (by zugzwang).
2) The auto draw function should not apply for centaur tours, as far as it is based on "insufficient material" (with the only exception K vs K).
It also shows, that a modified engine is able to evaluate the chances of stalemate, but it is still an open question, whether in other cases it might handicap itself as going for a stalemate, when it would still have chances to achieve a mate. Also it's not clear, how playing without tablebases affects the chances.
-
- Posts: 45
- Joined: Sat Jun 21, 2014 7:00 pm
Re: Request for engines with modified stalemate values
And another topical example from IC server:
Black's king is cut off. That's why Black brings his rook to the f-file and (according to "normal" draw rules) does not worry on exchanging rooks.
Yet, according to changed stalemate score he would not be happy to do so.. He would only be happy, not to lose a full point.
I think, it's only fair to give White here a stalemate bonus point, if he is able to force the K+P vs K ending. White would not be able to force rook exchange and stalemate, if Black's king would already be blocking the d-pawn (Kd6 or Kd5 instead of Kg7). In that position Black could actually reach a "normal" draw. And this would be fair too.
On the server "auto draw" applied after rook exchange.
Black's king is cut off. That's why Black brings his rook to the f-file and (according to "normal" draw rules) does not worry on exchanging rooks.
Yet, according to changed stalemate score he would not be happy to do so.. He would only be happy, not to lose a full point.
I think, it's only fair to give White here a stalemate bonus point, if he is able to force the K+P vs K ending. White would not be able to force rook exchange and stalemate, if Black's king would already be blocking the d-pawn (Kd6 or Kd5 instead of Kg7). In that position Black could actually reach a "normal" draw. And this would be fair too.
On the server "auto draw" applied after rook exchange.