Anything with low volumes isn't fair to compare, you can pretty much always pick something that "beats" the "market"
What is useful is to figure out what, today, may predict outperformance tomorrow. In choosing fund managers, it's not enough to choose managers who've outperformed in the past - that's not in itself predictive of the future (same is true for home prices as we learned in 2008, individual stocks, etc.). But other characteristics might be predictive, like choosing fund managers who follow a true "value" approach to stock-picking: finding companies with a competitive advantage and buying them below their intrinsic value being one approach to value.
See Warren Buffett's 1984 article "The Superinvestors of Graham and Doddsville" for the best example I know of distinguishing chance price rises from predictable rises: https://www8.gsb.columbia.edu/articles/columbia-business/sup...
It's not only not useful, it's totally obvious some things will always beat the market. "Beating the market" is really just "beating the average". If there's nothing that beats the average, then it means that all stocks do exactly the same thing...
But yes like you said finding things that beat the market in hindsight certainly isn't very useful unless you can learn some specific strategy (e.g. insider information) that was used and employ it yourself (if you choose).
What the did was take a group of college students and have them record what they did, what they ate, and various metrics like their daily weight.
When the “study” was over, the poured over the data looking for some correlation. In all that data, there was sure to be some random correlation, and they found one between chocolate and losing weight.
They then planted the story and watched everyone latch onto it. The follow-up story was about statistics and science, where they pointed out that given a small sample and a short time frame, and given that you are looking for a correlation, you will find one.
But confirming that correlation is where science begins, by testing the hypothesis with appropriate samples, time frames, and controls.