I upvoted for relevance (although it’s not quite what I asked for). Interesting read.
The efficient-market hypothesis is coherent, falsifiable and wrong, even in its weakest form: The paper describes a way to introduce an inefficiency, that were it arbitraged optimally, would solve an (NP-complete) satisfiability problem. Even if P = NP (which seems unlikely), it remains unproven, so nobody yet knows how to solve NP-complete problems in PTIME, which would be required for markets to be truly efficient.
The eldritch abomination may grow stronger with every attack, but it is decaying even faster than we can fight it. The number of possibilities to search for anomalies grows faster than our ability to compute them when you account for the possibility of anomalies in the pricing of sets of securities, like for pairs trading. There might be something like 100,000 publicly-traded stocks in the world, but that would mean (1000002) combinations or nearly five billion pairs. And that’s not even considering larger subsets, like triples, which would be on the order of hundreds of trillions.
So the EMH can’t be literally true. but can it be approximately true? It must be to some degree for the question to be this contentious. The real question I’m interested in, is “Are the markets exploitable?” Is there such a thing as alpha, or its it just luck? If finding exploits is as expensive as exploiting them is profitable, then the EMH might as well be true for that purpose. But if there are plenty of anomalies to go around, then I could have a realistic chance of finding and profiting from one nobody has noticed before.
A different paper but in the same vein: Markets are efficient if and only if P= NP
I upvoted for relevance (although it’s not quite what I asked for). Interesting read.
The efficient-market hypothesis is coherent, falsifiable and wrong, even in its weakest form: The paper describes a way to introduce an inefficiency, that were it arbitraged optimally, would solve an (NP-complete) satisfiability problem. Even if P = NP (which seems unlikely), it remains unproven, so nobody yet knows how to solve NP-complete problems in PTIME, which would be required for markets to be truly efficient.
The eldritch abomination may grow stronger with every attack, but it is decaying even faster than we can fight it. The number of possibilities to search for anomalies grows faster than our ability to compute them when you account for the possibility of anomalies in the pricing of sets of securities, like for pairs trading. There might be something like 100,000 publicly-traded stocks in the world, but that would mean (1000002) combinations or nearly five billion pairs. And that’s not even considering larger subsets, like triples, which would be on the order of hundreds of trillions.
So the EMH can’t be literally true. but can it be approximately true? It must be to some degree for the question to be this contentious. The real question I’m interested in, is “Are the markets exploitable?” Is there such a thing as alpha, or its it just luck? If finding exploits is as expensive as exploiting them is profitable, then the EMH might as well be true for that purpose. But if there are plenty of anomalies to go around, then I could have a realistic chance of finding and profiting from one nobody has noticed before.