I know it is not obvious to have a rich database of puzzles. The reason being that it requires a human touch. Automatically generating puzzles based on the existing game database using computer analysis does not always yield nice puzzles. Here are some examples:
Example 1: http://en.lichess.org/training/31299
There is several ways to mate in 3 but only one is accepted as correct. Why?
Win material, why does the last move matter?
Example 2: http://en.lichess.org/training/29081
Win a rook in one move - obvious! But then somehow the last move matters, like if a pawn will make a difference after that.
Example 3: Again bishop takes knight, but somehow the last moves is deemed important even though a requested computer analysis on the same game by lichess server suggests a different last move (Rb8 instead of Bg6)
After having spent a lot of time in the training section, it turned out to be very frustrating as the feeling one gets is that unless you evaluate each move as a computer, you will rarely succeed. But humans don't think as computers, and I don't believe training to do so should be the desired goal here.
Perhaps some sort of voting on puzzles should be involved before one tries to attempt the puzzles, like if 10 people who have a certain training rating vote for a puzzle, only then it becomes part of the training section, and before that it is in some sort of a "beta training" that does not affect rating, otherwise it's sort of frustrating as one goes through a large number of failures and feels unrewarded.
Example 1: http://en.lichess.org/training/31299
There is several ways to mate in 3 but only one is accepted as correct. Why?
Win material, why does the last move matter?
Example 2: http://en.lichess.org/training/29081
Win a rook in one move - obvious! But then somehow the last move matters, like if a pawn will make a difference after that.
Example 3: Again bishop takes knight, but somehow the last moves is deemed important even though a requested computer analysis on the same game by lichess server suggests a different last move (Rb8 instead of Bg6)
After having spent a lot of time in the training section, it turned out to be very frustrating as the feeling one gets is that unless you evaluate each move as a computer, you will rarely succeed. But humans don't think as computers, and I don't believe training to do so should be the desired goal here.
Perhaps some sort of voting on puzzles should be involved before one tries to attempt the puzzles, like if 10 people who have a certain training rating vote for a puzzle, only then it becomes part of the training section, and before that it is in some sort of a "beta training" that does not affect rating, otherwise it's sort of frustrating as one goes through a large number of failures and feels unrewarded.