lichess.org
Donate

The tactics trainer needs serious improvements

I just failed because I went for a mate in 2 instead of a mate in 1 (problem 41237). I get the point that It's about finding the best move, yet this is ridiculous, back to chesstempo.
A quicker mate can make all the difference in very fast games.
lol this is how puzzles works, you need to find most forceful and quickest way
otherwise for some positions you can check and drive the king back and forth 2 repetitions and delay the checkmate
or in an instanly winning middlegame position puzzle why would someone go for a winning endgame, chess is all about tempo, all about time
It is not so bad, sometimes it tells you that your move is good but you can do better. Nevertheless, it has happened to me that my move was actually better than the solution (and I went to see what Stockfish 8 would have played: it was my move and not the correction...).

In the game you describe, perhaps you have chosen a move that would have let the opponent play his queen in B8?
#1 I'm experimenting with both Ken Regan's IPR (intrinsic performance rating) and various search techniques (especially Proof-Number Search and its derivatives) in the hopes that it's feasible without overwhelming servers to improve Training.

The main problem is that it might be costly to "play against the machine" every time a user plays an unexpected move, so either Training needs to expect more moves or react differently instead of instantly failing the solver.
To my mind finding a clear Win (on material) should be enough in any tactics training.
"Mate in three" and such goals are for the problematists.

Still I think the tactics training is excellent. I use it a lot with my junior students.
Also, in most puzzles if you find a mate in N+1 instead of a mate in N, lichess will say "Good move but you can do better" and allow you to retry. I'm unsure why that didn't happen in this case.
I second the requested improvement. Finding a mate in N+x instead of a mate in N and still failing the puzzle has happened to me as well. The same is true for finding a, say, +5 eval solution instead of a +15.

Since there is already the "Good move.." function in place I think it's just not yet working well enough.
In the greatest game of the (last) century (you can find it from a link here, in the videos section) , Fischer did not deliver mate (in the final king hunt) in the smallest number of moves as possible, and who cares? But If that king hunt turns into a tactics problem here, Fischer`s last sequence would be considered a failure, or at least we would get a message "there`s something better" (yeah, for his opponent: resign!).
I agree the tactics trainer is a mess. There's too many problems where I play a move that is easily winning and +5 or more and I don't even get a "good move but you can do better message".

On problem 1186 http://en.lichess.org/training/1186 the last move, Qxd5 fails the problem and Bxd5 passes the problem even though both win a piece, and even though Qxd5 forces simplification at a time when, up a piece, the simplification is desirable.

I like a lot of things about the trainer and I use it b/c I do everything else here. But there seem to be a lot of kinks in it that don't make sense.

This topic has been archived and can no longer be replied to.