[asgard1024]

> If you want a reasonably efficient market, you need some participants to have close to perfect information.

Is this proven somewhere or you just assume the optimal strategy for markets is continuous?

I mean, it's not clear that the optimal strategy for "slightly imperfect markets" is at all close to the optimal strategy for markets with perfect information. And I actually doubt it can be proven, in the general case.

[ska]

Not only is it not clear, intuition from other areas of optimization would suggest it's unlikely to be true.

I've asked a couple of economists about this, but didn't get a satisfying answer. To be fair, it wasn't their area at all - and I may just have misunderstood what they were saying.

[tjgq]

I'm curious: can you name an optimization problem you would get that intuition from?

[asgard1024]

I am not a game theorist, but take Centipede game for instance. If you know the exact number of rounds in advance, the optimal strategy is markedly different than if you don't know it. And I think there are many weird behaviors like that in iterated games, where optimal solution for infinite time horizon is not the same (or doesn't even exist) as the limit of optimal solutions for finite horizons approaching infinity.