IlyaShpitser comments on Rationality Quotes December 2012 - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (226)
Although I would infer him to have already been aware of certain exceptions, e.g.:
If you and your opponent both know eachother to be skilled and masters of the way, then attempting the same tactic a third time becomes a different tactic; you are making an attack which your opponent would never expect you to try again.
It's a common failure mode of intermediate-level students of a competitive discipline to fall prey to the simplest tactics used by the weakest beginners, simply because they expect their opponent not to use said weak and simple tactics. For example, experienced chess players losing repeatedly to a complete beginner because they're used to playing against stronger and stronger opponents as they were gaining their experience. I've seen it happen often, and it's happened to me too.
ETA: I acknowledge that the example above isn't quite what is seen out there. What I had in mind was a one-off thing - of intermediate players, a "significant amount" (not the majority, but I'd guess "enough so that most experienced players have seen it happen more than once") go through a point at least once where they repeatedly lose against a player much less experienced than them, because of the reasons above. This second point is also debatable, but I think it's worth splitting and distinguishing between the two.
One interesting question here is what are the features of games where this does happen? In my view a much weaker player may win against a much stronger player in a game where the following features are present:
(a) There is strong "metagame" (that is, multiple equilibria). A beginner may not be aware of the current equilibrium and may defect in hard to predict ways that may give an advantage.
(b) There is randomness. A much stronger player may be modelled as a computationally omnipotent adversary. Such adversaries still cannot "read the minds" of randomness sources. A game that interprets a beginner's flailing as randomness can make it situationally powerful.
(c) The game is short enough such that the stronger player cannot learn what's going on due to (a) or (b) quickly enough to turn the tide.