[etpcxojp] Learning to beat the weak computer

The best move is the one which induces your opponent to make the biggest mistake.  If you are playing (practicing against) a deterministic computer program (e.g., chess) on a "weak" level, one does not need to speculate what the computer's response would be for each of your moves.  Just try them out.

Recalculate each computer move at a "strong" level to automatically determine whether a "weak" level move was bad.  The end result is learning to set traps, possibly risky traps that would backfire if you didn't know the opponent's response.  It might be a good way to study tactics, and to try to salvage a lost position.

Extend this idea from one-ply search to deeper.

(This is in contrast to Nash Equibrium or Minimax, which assumes the opponent will play the optimal move.)