Currently, in training puzzles, when you move a pawn to the back rank, it auto-promotes. But sometimes, a puzzle requires a clever underpromotion to a different piece. This can lead to unexpected behaviour where you think the pawn is going to auto-promote to a piece and it turns out to be different. It could also make puzzles easier because players who fail to spot the underpromotion could solve the puzzle by luck.
For example, http://en.lichess.org/training/37624 involves an underpromotion, but players expecting the pawn to turn into a queen may not see the right solution.
Changing the settings in preferences does not affect this, nor does logging out.
For example, http://en.lichess.org/training/37624 involves an underpromotion, but players expecting the pawn to turn into a queen may not see the right solution.
Changing the settings in preferences does not affect this, nor does logging out.