Hacking the random walk hypothesis

It's worth noting that its pretty easy to generate data that passes all NIST random tests which is 100% predictable / pseudo random. Eg, just encrypt using AES 256 with the key "predictable" hashed. So, these tests are necessary but not sufficient to prove a random walk hypothesis. The author says as much, but buries this somewhere midway down the article.

> Markets are, quite simply, not random.

In the strict statistical sense, no one has ever disputed this... EMH is about excess returns. Of course the market has long-term trends - they go up, because the world is getting richer. Specifically, they go up something like 7% annually in the USA. So if you work with daily returns, your randomness suite had better reject the hypothesis of randomness!

I've always assumed that the randomness of the markets was self-healing: if the markets exhibit non-randomness (and thus become predictable) some arbitrageur will step in and correct it and then some trader will figure out a way to make money against the arbitrage. The best analogy I know of is Rock-Paper-Scissors. The commonly accepted optimal strategy is random, but against a random strategy always choosing rock works exactly as well. It is only if your opponent recognizes the lack of randomness and breaks their own randomness that they can capitalize, at which point you can counter.

How on earth does TFA relate to the Guy-Fawkes-mask-with-star-spangled-bandanna look in the accompanying picture?

I wonder if there's a timescale-dependence of the appearance of randomness here. Subject the data to a series of Gaussian high bandpass filters in Fourier space and repeat the analysis. Does the randomness get higher the higher the frequency?