There was an article a while back -- https://news.ycombinator.com/item?id=1144548 -- purporting that the efficient market hypothesis is true if and only if P = NP. ... which I'd argue implies social science is overly optimistic about that hypothesis.
That also implies that if the market is very close to efficient, you need absolutely massive amounts of computational power to bring it even closer (i.e. to make money). So even if it is inefficient, that doesn't mean anyone can reasonably expect to find exploitable patterns.
