[jjrxvbom] Learn to make chess mistakes

Use machine learning techniques on previous human games to predict whether or not a human will make a mistake (blunder) in a given position (time information might be nice), and what kind of mistake they will make. Features may be extracted from the game tree from the position.

Create a more "human-like" chess program, a trainer for playing against humans, or a particular human.

Not only learn to induce the opponent to make mistakes, but set the computer to model yourself and learn to make suboptimal but "solid" moves which avoid you getting into situations where you make mistakes.

Expectimax.