POLL: What is more important?

[Marek]

A perfect strategy is one in which chess is solved and lays out all the moves to make in order to win during positions.

What does a "perfect" strategy recommend in a drawn or lost position? Toss a coin?

[xshat]

A perfect strategy is one in which chess is solved and lays out all the moves to make in order to win during positions.

What does a "perfect" strategy recommend in a drawn or lost position? Toss a coin?

I cannot say because chess has not been solved yet so we do not know.

[Marek]

A perfect strategy is one in which chess is solved and lays out all the moves to make in order to win during positions.

What does a "perfect" strategy recommend in a drawn or lost position? Toss a coin?

I cannot say because chess has not been solved yet so we do not know.

Another thing we do not know is what opposition a "perfect" engine will face. Which leads to an indissoluble problem. In a losing position against a tactical engine one strategy will score best in practice: in a losing position against a positional engine a different strategy will score best in practice. Yes, there will be an optimal strategy for each different circumstance, but not a perfect strategy for every circumstance.

32 man tablebases are practically impossible, but a perfect strategy is logically impossible.

[Uly]

Yes, I finally see it, xshat thinks that there's some strategy to win from drawn positions against imperfect opposition, that would make all perfect engines to play the same. That assumes that the moves that would beat Rybka 4 would also beat Kasparov, and me, but that's illogical! The position is called "drawn" for a reason, any non perfect opposition could play the perfect moves by chance and draw.

A simple example could be made: Create a randomizer engine that always chooses different moves on future games (Rybka Randomizer already does work for this), and play enough games between the "perfect strategy" and the randomizer, eventually the randomizer will get a draw, which means such perfect strategy cannot exist.

What moves would a perfect strategy play against another perfect strategy? If all engines with perfected chess play the same against each other, it's easy to see what moves they are on a single game, copy them and become "perfect", so it's obvious an optimal strategy would play differently each game to avoid this problem, and then, chess engines with perfected chess would play differently from one another.

[hyatt]

I don't follow any of the above. For any position, one of the three following ideas is true.

(1) side to move wins with perfect play. Doesn't matter what the opponent does, the side to move will win, since it will never play a non-perfect move.

(2) side to move loses with perfect play. In this case, the opponent can play a poor move and convert our loss into a draw or win, but if we assume perfect play by both sides, side to move loses.

(3) position is a draw with perfect play. Obviously either side can make a blunder and lose, but with perfect play this ends in a draw. All the other stuff is pretty much irrelevant.

When you get into perfect play against an imperfect opponent, things can change a bit, because you might not want to play perfectly in a drawn position if your opponent is not perfect. You'd prefer to play in a way that gives your opponent the most opportunities to go wrong, rather than in the way that leads to the longest (or shortest) draw. But this is a completely different topic...

[Uly]

But this is a completely different topic...

It is related, because xshat argues that the engines' origins don't matter, because when chess is solved all engines will play the same. Your third point means that when chess is solved, engines will focus in going to positions that will increase the chances of the opponent blundering, but there are many ways of going about it, then, all engines will play differently when chess is solved, so their origin will matter even at that point, thus, their origin matters now.

[xshat]

A perfect strategy is one in which chess is solved and lays out all the moves to make in order to win during positions.

What does a "perfect" strategy recommend in a drawn or lost position? Toss a coin?

I cannot say because chess has not been solved yet so we do not know.

Another thing we do not know is what opposition a "perfect" engine will face. Which leads to an indissoluble problem. In a losing position against a tactical engine one strategy will score best in practice: in a losing position against a positional engine a different strategy will score best in practice. Yes, there will be an optimal strategy for each different circumstance, but not a perfect strategy for every circumstance.

32 man tablebases are practically impossible, but a perfect strategy is logically impossible.

There could be a perfect strategy for every circumstance, we will not know until chess is solved.

[xshat]

Yes, I finally see it, xshat thinks that there's some strategy to win from drawn positions against imperfect opposition, that would make all perfect engines to play the same. That assumes that the moves that would beat Rybka 4 would also beat Kasparov, and me, but that's illogical! The position is called "drawn" for a reason, any non perfect opposition could play the perfect moves by chance and draw.

A simple example could be made: Create a randomizer engine that always chooses different moves on future games (Rybka Randomizer already does work for this), and play enough games between the "perfect strategy" and the randomizer, eventually the randomizer will get a draw, which means such perfect strategy cannot exist.

What moves would a perfect strategy play against another perfect strategy? If all engines with perfected chess play the same against each other, it's easy to see what moves they are on a single game, copy them and become "perfect", so it's obvious an optimal strategy would play differently each game to avoid this problem, and then, chess engines with perfected chess would play differently from one another.

The very best strategy would be utilized by the very best engine in any situation, winning losing or drawn, until the game ended.

[xshat]

But this is a completely different topic...

It is related, because xshat argues that the engines' origins don't matter, because when chess is solved all engines will play the same. Your third point means that when chess is solved, engines will focus in going to positions that will increase the chances of the opponent blundering, but there are many ways of going about it, then, all engines will play differently when chess is solved, so their origin will matter even at that point, thus, their origin matters now.

That's a lie, I never once said an engine's origins don't matter. I said when chess is solved engines will play the same, which is NOT saying engine's origins don't matter. I find it strange you keep putting words in my mouth I've never said. You think all engines will play differently when chess is solved, that's like saying the checkers engines which have been programmed to solve checkers are going to play differently. Well, they don't.

[Marek]

I don't follow any of the above. For any position, one of the three following ideas is true.

(1) side to move wins with perfect play. Doesn't matter what the opponent does, the side to move will win, since it will never play a non-perfect move.

(2) side to move loses with perfect play. In this case, the opponent can play a poor move and convert our loss into a draw or win, but if we assume perfect play by both sides, side to move loses.

(3) position is a draw with perfect play. Obviously either side can make a blunder and lose, but with perfect play this ends in a draw. All the other stuff is pretty much irrelevant.

When you get into perfect play against an imperfect opponent, things can change a bit, because you might not want to play perfectly in a drawn position if your opponent is not perfect. You'd prefer to play in a way that gives your opponent the most opportunities to go wrong, rather than in the way that leads to the longest (or shortest) draw. But this is a completely different topic...

(1) is right of course.
(2) The world is populated by all kinds of chessplayers. We can say what will happen if there is perfect play: we cannot say that there will be perfect play.
(3) Again, we cannot say that there will be perfect play. You say, "All the other stuff is pretty much irrelevant." Irrelevant to what?

The OP (xshat) started this very thread with, "Chess is chess, and when the day comes that chess is solved, all chess engines will be clones of each other." Those were his opening words. I relevantly challenged this assertion, and painstakingly argued against it with some success. The success I claim was when the OP made the concession through the comment "... programmed to have solved chess, which includes coding outside of the 32men bases." Presumably the "outside" coding was to cover play in lost or drawn positions. (There are further arguments in this thread as to why this outside coding will be different for each engine and impossible to perfect.)

There is confusion throughout over the terms perfect and solved. Obviously game theory/mathematics has its own technical definitions which must be respected in its fields. For chessplayers (including engines of course) the continual question is what move to play next. Even when chess is formally solved (with 32TBs) players will still be wondering what to do in a lost or drawn position against an unknown opponent. Paradoxically, we'll have the solution but not the answer to our question.