Another factor is that the weak inefficient market hypothesis (as far as I know) isn't a claim that we can't find the method for analyzing past prices. It's a claim that there is no possible method because the future price of a security is for all purposes stochastically independent of past prices. No matter what algorithm you use, NP, P, or any other algorithmic space, you can never reliably predict the value of a variable that is independent of its previous values.

I'm not positive that this defeats the argument this guy was making (not a computer scientist, only a programmer), but from what I could gather it seems to.