PZ170320

White has just moved his bishop from d5 to f7. What is the best reply for this ? 

It's a good combination but black has an equal choice between 5. .. Qf2 and 5. .. Qf4. So it is a flawed puzzle in this form. Better is the line 5. Bd5 which forces black to play the only clearly winning move 5. .. Qxd2! A puzzle remains correct as long as there is just one option for the 'leader'.

All clearly losing moves are equal.

The quantative logic of defining as "the best move" the one that offers the longest resistance or loses the least amount of material does not apply to problems or puzzles in general. Except of course when quantification is stipulated as in direct mates.

What does count is the qualitative judgement of which lines are hardest to find/refute by a human being or the most aesthetic or reflect the ideas of a composer or game player best. Of course a quantitative line may coincide with a qualitative line but that really does not add any value to the line.

The proof of this can be easily found in many games and endgames. Isn't it better to risk a checkmate that is hard to see than go 2 pawns down and lose for sure? All GMs are nodding. Isn't it ridiculous to escape into a long tablebase loss when the composer of the endgame has prepared a beautiful trap for us?

Engine scoring by minimax-logic does not exist in the world of puzzling, though it has tremendous value to the analysts who have different goals.

It is arguable that a game fragment does not demand the same standards as another puzzle or composition but there are three issues with that: (a) the chess.com interface does not support alternative correct moves; as a solver you are red-flagged when you make the alt choice which is unjust; I complained many times that chess.com should improve it's interface on this point (b) There is a game forum; if a game fragment is not a decent puzzle, then you can place it there (c) often the dual can be removed by making a small change to the position; most posters however refrain from making those changes since they are only 'showing off games' and not presenting us with a puzzle.

Thus, the unique sequence is enforced by the chess.com interface. Respect for the puzzle solver requires that you abide by it when presenting a puzzle.

This is a summary of discussions taking place on the 'Daily Puzzle' over time. Currently, the DP-staff seems to have the balance right between 'what is peceived as the best defense' and 'which unambiguous line is picked for the solution'. It's not math, it's human intuition.

Hi @Arisktotle. I agree it requires judgment. The only reason for changing the default play here is the crappy chess.com interface. Otherwise, from a pedagogic point of view, Qf2/Qf4 should both accepted as part of the main line here. QRNvQR is a very different proposition to RNvB, even though there are pawns to left and right in both, and if I were White I would definitely steer to the former, to improve the pay-off.

Thanks for complaining energetically to the powers that be about the interface.

Actually, I would not play your way but take the chance that my opponent would not see .. Qxd2. That chance might look small but it is not smaller than the chance that a player would not win the ending with a piece up + an overwhelming position!

In general, all puzzle and tactics logic based on objective quantification fails on obvious grounds. For instance, many tactics start with innocent moves like 'exhanging a piece' or 'attacking a piece with a pawn'. By your argument, the appointed victim should escape into giving up a piece at that point since the alternative proves to be much worse (3 moves later). But nobody does that since nobody ever plays a move that clearly loses. Instead, every decent player challenges his opponent to prove his point by clearing the still present hurdles. If you were right, probably 50% of all tactical puzzles are incorrect!

John Nunn has made some interesting comment on the superstition that 'the best move' is 'the move that wins quickest' and that GMs supposedly go for that. GMs, just like normal chess players go for 'the most certain win' which is the one that requires the clearing of the minimum amount of hurdles - relative to player strength. By implication, the task of the defense is opposite to that. Not trying to delay defeat but instead place the toughest obstacles on the road they can, even if that (possibly) leads to a quick defeat.

Note that the tactical strike .. Qxd2 is a one-time opportunity; when you miss it, it is gone. On the other hand, in the endgame with a piece up you must stumble a dozen or so times before throwing away the win. That's what I call a hurdle vs no-hurdle.

Certainly over the board, these kinds of psychological tricks are important, but I guess you are accepting my point that in a tactical puzzle, for pedagogical purposes, it's important that the defence give it their best objective shot.

I don't think we do agree on that on this level but probably on a higher level . The idea of chess problems or tactics or just chess playing in general is that one should calculate all the relevant variations. The 'solution' is not defined by one presumed 'best defense' but by the knowledge that all escape routes for the opponent are blocked. The pedagogical method should therefore focus on testing that 'the subject' did understand the complete challenge put to him. In the current problem I would think that a solver would need to show the replies to both Bd5 and Rd3. There is no pedagogical preference between them because only 'complete understanding' is acceptable.

I do not know of any problem-solving interface capable of handling multi-thread solutions, but I am sure one will exist one day. Until that time I am quite happy to permit puzzle posters to make their own random defense selection, provided (a) it demonstrates relevant content (b) the puzzle interface can handle it, e.g. with regard to ambiguities.

And it will be a long time indeed before problems involving retraction are handled here. Let alone ones with both forward and backward play! But the main task is not a programming one, it's an extension to PGN. Maybe PDB is a good start, but curiously Gerd Wilts has not been made his notation fully public.