But as you say about chess, I'm not sure if this is a very generalizable idea, at least when it comes to rationality.

For most practical situations I would suggest that it does generalise. Humans have relatively little ability to compound on success in a drastic manner. Exceptions of course include situations such as if Smily and Clippy were created at the same time on the same planet. Clippy getting the first week wrong could well leave tiling the universe with paperclips instead of molecular smiley faces is completely beyond his grasp.