wgarvin wrote:... if it says something like "win in 53" the engine should treat that as a draw.
To treat it as a draw would e a bad mistake. It is only a draw against nearly perfect defense. If you treat it like a draw, the engine wouldn't even make an effort, and just as likely makes the first move to a position that is an immediate stalemate.
It is like playing KBPPKB, which is often drawable with unlike Bishops. Nevertheless sacrificing both your Pawns and your Bishop in the first 3 moves is bad strategy, despite the fact that KKB is also still a draw.
You want the engine to play as strong as it can. You don't want it to squander the advantage it has, just because it was not yet big enough for a certain win.
wgarvin wrote:... if it says something like "win in 53" the engine should treat that as a draw.
To treat it as a draw would e a bad mistake. It is only a draw against nearly perfect defense. If you treat it like a draw, the engine wouldn't even make an effort, and just as likely makes the first move to a position that is an immediate stalemate.
It is like playing KBPPKB, which is often drawable with unlike Bishops. Nevertheless sacrificing both your Pawns and your Bishop in the first 3 moves is bad strategy, despite the fact that KKB is also still a draw.
You want the engine to play as strong as it can. You don't want it to squander the advantage it has, just because it was not yet big enough for a certain win.
I think that there is the perfect place for "swindle mode" as defined in Crafty. It would do that perfectly.
wgarvin wrote:... if it says something like "win in 53" the engine should treat that as a draw.
To treat it as a draw would e a bad mistake. It is only a draw against nearly perfect defense. If you treat it like a draw, the engine wouldn't even make an effort, and just as likely makes the first move to a position that is an immediate stalemate.
It is like playing KBPPKB, which is often drawable with unlike Bishops. Nevertheless sacrificing both your Pawns and your Bishop in the first 3 moves is bad strategy, despite the fact that KKB is also still a draw.
You want the engine to play as strong as it can. You don't want it to squander the advantage it has, just because it was not yet big enough for a certain win.
That cuts both ways though. If you treat a position as a "win in N" even though the opponent may actually be able to draw it, you might give away an actual advantage in order to pursue this unachievable win.
If the engine can't force a win against an optimal opponent, its not really a "won" position. At best its a "draw that some opponents have a hard time holding onto". I agree that you want to try and put up a good fight if you find yourself in such a position (like Crafty's "swindle" mode does). If you're lucky enough to be playing against a fallible opponent, they might make a sub-optimal decision and it might suddenly become a clear win for you.
However, I don't think you want the engine to search towards such a position, thinking it is "won". You especially don't want it to play out a "winning" line from a database and then convert to such a position and find that it can no longer win. Thats why a converting move out of a DT?50 database, which goes to such a position, is treated as a drawing move rather than a winning move--better safe than sorry!
wgarvin wrote:... if it says something like "win in 53" the engine should treat that as a draw.
To treat it as a draw would e a bad mistake. It is only a draw against nearly perfect defense. If you treat it like a draw, the engine wouldn't even make an effort, and just as likely makes the first move to a position that is an immediate stalemate.
It is like playing KBPPKB, which is often drawable with unlike Bishops. Nevertheless sacrificing both your Pawns and your Bishop in the first 3 moves is bad strategy, despite the fact that KKB is also still a draw.
You want the engine to play as strong as it can. You don't want it to squander the advantage it has, just because it was not yet big enough for a certain win.
That cuts both ways though. If you treat a position as a "win in N" even though the opponent may actually be able to draw it, you might give away an actual advantage in order to pursue this unachievable win.
If the engine can't force a win against an optimal opponent, its not really a "won" position. At best its a "draw that some opponents have a hard time holding onto". I agree that you want to try and put up a good fight if you find yourself in such a position (like Crafty's "swindle" mode does). If you're lucky enough to be playing against a fallible opponent, they might make a sub-optimal decision and it might suddenly become a clear win for you.
However, I don't think you want the engine to search towards such a position, thinking it is "won". You especially don't want it to play out a "winning" line from a database and then convert to such a position and find that it can no longer win. Thats why a converting move out of a DT?50 database, which goes to such a position, is treated as a drawing move rather than a winning move--better safe than sorry!
A position that is Mate in 53 (draw under the 50 move rule) should not be considered a draw or a mate. Maybe half a pawn or so (the amount is debatable). The advantage is real because the opponent is forced to make 50 perfect moves with only two chances to err.
wgarvin wrote:... if it says something like "win in 53" the engine should treat that as a draw.
To treat it as a draw would e a bad mistake. It is only a draw against nearly perfect defense. If you treat it like a draw, the engine wouldn't even make an effort, and just as likely makes the first move to a position that is an immediate stalemate.
It is like playing KBPPKB, which is often drawable with unlike Bishops. Nevertheless sacrificing both your Pawns and your Bishop in the first 3 moves is bad strategy, despite the fact that KKB is also still a draw.
You want the engine to play as strong as it can. You don't want it to squander the advantage it has, just because it was not yet big enough for a certain win.
That cuts both ways though. If you treat a position as a "win in N" even though the opponent may actually be able to draw it, you might give away an actual advantage in order to pursue this unachievable win.
If the engine can't force a win against an optimal opponent, its not really a "won" position. At best its a "draw that some opponents have a hard time holding onto". I agree that you want to try and put up a good fight if you find yourself in such a position (like Crafty's "swindle" mode does). If you're lucky enough to be playing against a fallible opponent, they might make a sub-optimal decision and it might suddenly become a clear win for you.
However, I don't think you want the engine to search towards such a position, thinking it is "won". You especially don't want it to play out a "winning" line from a database and then convert to such a position and find that it can no longer win. Thats why a converting move out of a DT?50 database, which goes to such a position, is treated as a drawing move rather than a winning move--better safe than sorry!
A position that is Mate in 53 (draw under the 50 move rule) should not be considered a draw or a mate. Maybe half a pawn or so (the amount is debatable). The advantage is real because the opponent is forced to make 50 perfect moves with only two chances to err.
Miguel
I do agree with you, and with hgm. If your opponent has the right kind of tablebases, you can't actually win that position against him. But its still a better position to be in than a simple draw, and against many opponents you might still be able to win it. I just don't think you want to direct the search towards such positions, unless that's all it can find.
Any position where the engine can force a win no matter what, should score more highly than these positions where it needs a bit of "help" from the opponent in order to win. It seems safer to treat a DTZ50 result saying "win in 53" as if it was "a little bit better than a draw" rather than "a little bit worse than a forced win".
"A little bit" is actually not that little. The draw margin, if you are not yet in a Pawn ending, is usually > 1 Pawn. For a DTZ=51 position, which is just on the brink of a win, a score of +1.5 would be natural. Preferring non-tablebase positions with +1 or even +0.5 over it will not increase the win chaces anymore than refraining from capturing a Pawn in KNKP to avoid the KNK draw will.
