Ideal trainer versus standard puzzles

A major drawback with current puzzles is that you know that the correct sequence will lead to a win or decisive advantage. In a real game nobody tells you in advance that the position before you has that possibility. Moreover, in the entirety of game situations, it is rare that the best move wins. Often the best move creates a small strategic advantage, or successfully parries a threat, or even forces a draw with an otherwise losing position, or consolidates one's position by better coordinating pieces or opening a file, etc. The ideal "puzzle" trainer would randomly throw you all of these. Sometimes the best move would be a banal but solid positional move, possibly of mere defensive value at the moment. And when you CAN win, you don't know it in advance, so it takes greater skill or knowledge to see it. I wish there were a puzzle trainer of that sort. With the tool of computer analysis it should not be difficult to add these sorts of "puzzles" as well, and they can be genuine puzzles in the sense of not being obvious and requiring skill or knowledge to solve.

This is pretty old.

You can do it yourself, it is called random puzzling: open a random master game, jump in the middle of it and search the best move. :)

(using ChessBase you can set the notation to training which means the follow-up is hidden.)

Thanks for the suggestions. Correct me if I am wrong, but when you replay a master game, some of the moves are mistakes or blunders. If you doubt that, then recall that in one of the games that the great Bobby Fischer lost to Spassky in their renowned tournament, Fischer made the infantile blunder of taking a corner pawn with his bishop and having the bishop trapped with a pawn advance. The great Akiva Rubinstein lost his queen in a smother trap often enough (twice in tournaments) that the trap is named after him (a dubious honour). It's in the Q4 opening. The leading Japanese GM - I forget his name - lost a pawn in an opening Sicilian trap (vs. Carlsonn) that even a mediocre player like myself has used. Perhaps one could use as one's data base games between the strongest engines.

But I don't necessarily mean one should see just randomly selected positions. Some patterns are particularly instructive, even if only defensive or strategic. It is very useful, for instance, to know how to force a draw with an otherwise lost position.

You should not believe, that the computer knows, what the very best move in a randomly chosen position is. Computers are very good in not overlooking material winning moves, but they are far from perfect, when it is about strategical advances.

So, no, there won't be such a training tool.

@sheckley666 I believe your information might be outdated by 15 years. These days computers are stronger than humans in like 95% of the positions if not more. (yes, back when Fritz was called "strong" your statement was true, but look at Stockfish, Komodo or for that matter Leela)

@nayf I think on chesstempo there is a mode where you sometimes get position where the best move for you doesn't win but e.g. only hold equality or so. As for minor advantages that is difficult to decide (for humans or computers alike), it might be clear what the best move is but not so clear how much better than the second best move it really is. Plus different players have different styles so they might even disagree on what that best move is.

However including puzzles where you try not to lose rather than win is possible and I agree that it would be an improvement.

I tried building up my tactical skills with puzzles recently. Some of these puzzles I don't pass because they are the dumbest exchanges ever. At the very least the puzzles should come with a description of why that position is threatening...

@MoistvonLipwig Of course today's computers are stronger than humans in nearly any position, but not because of their better strategical understanding, but because they find tactics even, if the tactics are hidden behind several quiet moves.

Another thing is, that a human won't learn anything, if (s)he is shown the best (if the trainer knew it) move without any explanation, why this is the best move.

The explanation part is true, that also is a general problem even for tactical puzzles. (as the computer considers different lines "critical" than the human does)

As for the computer only seeing tactics even if they are behind quiet moves, that is true however that is exactly what strategy is. A move is strong strategically not because it somehow makes you magically win the game but rather because far in the future (beyond the humans tactical sight) it will make some tactics work. It's just that computers often (not always) can directly calculate that whereas the humans are too weak for that so they have to fall back to mere heuristics to estimate if a move helps in the future. (which then is called "strategic play")

Again I'll plug Chess Tempo's excellent puzzles:
beta.chesstempo.com/

Toadofsky, thanks for the suggestion; I'll check out Chess Tempo and report back. I used to do puzzles there, but they were all standard ones, at least under the setting I was using.
MoistvonLipwig, just to reinforce your point: the best engines today have ratings around 3200-3300, whereas the highest rating a GM ever had was in the 2800s. But yes, explanations are often necessary, even for GM games and occasionally standard puzzles. I prefer to watch youtube videos of games with commentary rather than play them out myself.
Just to clear up a misconception: some think that the strategic and defensive puzzles I would like to see added to the mix with tactical puzzles will have only small advantages that good players with different "styles" might decline to play. But those are NOT the ones that should be added. There are defensive moves that prevent you from losing! There are strategic moves which any player ought to make regardless of "style". And there are small tactical moves, say that capture only a pawn, which are also nevertheless moves any player ought to make, but don't show up in standard puzzles. (If you see a move that will capture only a pawn, you know in advance that that is the "wrong" solution in standard puzzles).
MoistvonLipwig, the interesting implication of your last remark is that to an ideal player with a super-engine-like mind, there would be no distinction between tactics and strategy. It is only human limitations which make the distinction possible. Actually that is not quite true. Strategies, as you yourself suggest, increase the LIKELIHOOD of tactical opportunities, but do not NECESSARILY lead to any tactic because the opposing player might prevent their arising. In other words, strategies increase chances in the future but are indeterminate because of the multiplicity of possible moves, whereas tactics FORCE a realised advantage.