Training, more tolerance with movements

I'm having fun doing and studing training chalenges, they are so great to learn and improve our chess vision.

However, sometimes I'm missing more tolerance at the moment to find the demanded continuation;
to improve my way of chess, what I'm doing is to make a fast view and try to find as fast as posible the solution, or at least, one of the best/great movement. I agree than sometimes its not the better movement, but I think than it should have to allow more tolerance and don't losing as many points if you were wrong when actually you were not as much.
Sometimes, actually, the system is already working and its giving you another option to re-try: "Good movement, but you can do it better" (or something like that, I have it in spanish)

So, I propose than you add this option to avoid losing the chalenge, and also your points.

Some examples I have where I was sad than they count me it bad, took my points, and don't allow me to success the chalenge:
Correct movement: (check mate in 2)
Incorrect movement: (winning for free the Queen in 2)
I agree than the check mate is better, but in the situation we were, we had both same pieces, and the end was my queen, bishop and rock against one Rock and Bishop, I would allow the "mistake" and confusion of thinking than capturing the Queen is the better movement. And at least, the program say you: "Good movement, but not the best one"

I also remember another one which the correct movements were checkmate in 7; when I was at correct movent 3; I decided to do another line (which was considered wrong, made me lose points, and it didn't allow me to win the points I deserved) My other line was: winning a Rock, forcing the contrary to sacrifice Bishop, Knight, and the other Rock, and eventually, it was checkmate in 11. (The truth is than I didnt see the checkmate in 11, but I saw than I could win many material so I was clearly going for that line)
I agree than the computer and the training wants me to find the best combination, but in that case, my combination was great enought and it didnt even say: "good but not the best" it simply take away my points.

What you think?

(as I'm recording videos during playing) I could create the last game I talked about:
That was the correct continuation:
http://en.lichess.org/1e97LBaW/black

That was my line which make me loss the points:
http://en.lichess.org/60eqtzJG/black

Thanks for your attention.

I agree with you. Maybe they could cut less points if we did about 4 right moves out of 5 than if we had missed the first move for example, just because losing a lot of points when you're not totally wrong is frustrating. I'm almost sure that there are lots of topics regarding this situation here. Although I don't know if they plan to do something about that.

"[...] what I'm doing is to make a fast view and try to find as fast as posible [possible] the solution, or at least, one of the best/great movement [...]"

Don't do that. If you don't have a complete and best follow up, you're missing the point of tactics training. Existing puzzle tolerance is already high enough.

Yeah, I 100% disagree with this idea. The example OP uses is a prime example of the completely flawed logic with this request. If you have mate in 2 and you instead grab material then you are giving up an absolute win for a risky sideline out of material greed -- the ultimate sin against playing winning chess, which is the entire purpose of these puzzles -- and your opponent now has counterplay (inaccuracy or mistake). Even worse, maybe you even missed that by grabbing that queen he has a tactic to force the material back to equal and actually have an advantage (blunder).

The training is exactly as it should be. People are daily posting complaints about this sort of thing or "incorrect" puzzle solutions only to be shown that they are mistaken. If you go for a move that lowers the evaluation of the position significantly, you fail the puzzle. If you go for a suboptimal move that doesn't make you lose more than a couple centipawns of your evaluation, you get a "good move, but you can do better" and you try again. There are very few cases (like the one being discussed elsewhere in this forum) where the engine teeters between 2 different lines and will usually choose the "good move" line over the solution. But generally these puzzles are not subjective, and the point is not to race them but rather to find the accurate move.

It sounds like what you'd like though would be a time attack variant for the puzzles, in which you are given maybe 2 or 3 chances to fail before you lose points and in which you also lose if you fail to solve the puzzle in a set time. THAT I think would be an AWESOME added feature, and THAT idea I would support 100%. Just don't mess with the training that we already have because it's fine.

In general such puzzles should have one clear solution. If you see that you can win a queen for nothing, you just lose interest. No need to seek for a more complicated line to a mate. The road to mate may be quicker, counted in moves. But in practice, if you win a queen, you can play very quickly, you don't need to calculate much anymore. It's often the better decision for a practical player. These are just engine analysis, so you have to consider it like this. Find the best move according to an engine, no human element. It's different from most puzzle books. But's it's also an interesting form of training. Play like an engine! And don't care so much about this rating. You can judge by yourself if you found a way to a clear win. Sometimes there is more than one, but only one is accepted. Check it with your computer and if you found a way to +12, while the official solution gives +12,2, don't care! There is no competition for best solvers, no ranks, it's just for you.
Muchos recuerdos!

I disagree with #5. I have had some very tough games where I went for the "practical" and "easy" win by grabbing the queen only to find my own position compromised and extremely challenging when it turned out I had a forced mate. And as I said, if the solution is +12.2 and your move is +12 you will get a "good move, but you can do better." You only fail when you make an inaccuracy, mistake, or blunder. Your argument has holes.

That said, you are absolutely correct that this is not the traditional means of tactics training and it is indeed more like training yourself to play like an engine in some ways. But I think the majority of these puzzles are selected in such a way that there is, in theory, only one solution. I refute people's "better" solutions to these puzzles all the time without the use of an engine to find why the selected line is better. And as it has been stated 100 times over, if you are playing under time pressure, finding a fast mate over finding a long road to a win can absolutely mean the difference between winning and losing the game. Even a mate in 3 vs. a mate in 4 can mean all the difference if you have only 5 seconds on your clock.

The lower-rated (non-checkmate) puzzle solutions sometimes leave the "winning" side with (dubious) temporary material gain but in such a flimsy position that an average player like me would find it very challenging to continue. (e.g. http://en.lichess.org/training/49473 , http://en.lichess.org/training/52154 , http://en.lichess.org/training/32107 ; for http://en.lichess.org/training/8592 the AI8 will work it differently from the puzzle line if you start the board from the puzzle's initial setting wherein you will be left with less material gain in a more compromising position.)

On another note, some puzzles lead to a draw (or at least I think they do; I try them myself several times against stockfish level 8 ... anonymously) which I guess is a "win" in the sense that you at least avoided losing a seemingly lost game. (e.g. http://en.lichess.org/training/9409 ; I encountered a few more, but I only recently started bookmarking these.)

By the way, I'm not actually complaining. I'm just sharing my experience and observations. (I might be wrong about some of these but then someone might enlighten me and I will have learned something new.) What I read and solved from books many, many years ago were all mating problems so having these mixed in gives it a refreshing twist. However, unlike checkmates which clearly spell the end of games, I'm hypothesizing that these material puzzles have the potential to be expanded to some kind of further training.

Indeed, sometimes playing to equalize after an opponent's mistake is the best you can do. I think they make some of the best puzzles because we often miss chances to equalize lost positions after our opponent blunders. I think puzzle 9409 is a great example, thanks for sharing that. It does indeed look rather drawish.

As for the puzzles listed you listed, let's see how dubious the final positions are...

49473: the plan is fairly straight forward after the final position. Check with the queen, it improves her position and Black has hanging pieces everywhere in this puzzle. If the king moves you probably castle to remove Black's counterplay and are still threatening to win a pawn to fork the knight and rook, so you know what Black will have to do next and plan from there. If the bishop comes back to block you win the knight with the rook and take away the threat of losing it to the pawn advance, then just castle. It seems a rather simple position to play from for White, hardly dubious at all.

52154: I'm guessing here you just take the pawn on f4 and you should have a winning endgame, being up a knight and a pawn. I would hardly call it dubious, but it is an endgame, and endgames are tricky things for beginners. But it's a winning endgame transition for sure.

32107: I'll agree, this one definitely looks complicated. My gut says just sacrifice the exchange with Rxc4 and then sidestep the king to safety on f7 and it should be easy enough for the queen and rook to coordinate with those pawns to rip open the a and b files to get to the king, so for an aggressive player this seems pretty straight forward of an idea and I doubt it's losing. And if you want to be more positional, there's just pulling the queen back to e3 and start forcing White to weaken his position in order to defend all those unprotected pawns anchoring the 4th rank. This second plan makes for a pretty straight forward plan and tactics will eventually arise once you get the rooks connected, so I'd once again say this position is not dubious as the hanging pawns are easy to spot as the next weakness to exploit.

8592: You've won the queen for a rook and your queen is on the long diagonal closest to the king that was just forced to move and now can't castle. and you still have the bishop pair in a very open game and a giant hole on g2 right next to the king. If this is dubious because it's too "complicated" to figure out what to do from here, then maybe it's time to take up a new hobby. ;)

Great examples of your thought, and I'll admit they aren't necessarily as simple as finishing a puzzle knowing you are in a forced mating line that wins or just getting mate to end the puzzle, but definitely not dubious final positions I think. 8592 even forces you to play the first natural move after winning the material to get the plan started. Somehow I failed all of those puzzles, though, so there goes my rating! :(

I agree with you people about the fact than one line can change completly the game. But c'mon, in the example I showed you, the mate was in 7 and I made mate in 11; and it was counted bad. Here there is no line of mistake its just "not the best checkmate sequence". But still, is a checkmate sequence.

addressing #5: unlike some chess sites which handpicks the best puzzles, lichess's puzzles are automatically generated. This means they are not as good as we hope they can all be. As a result, the popularity system is used instead for making good and clear puzzles float to the top.