The problem with the data mining approach is that it works spectacularly until it doesn't, and when it fails, it also fails spectacularly.

That's true of any strategy in any situation in which your strategy affects the behavior of whatever system you're interacting with.

i.e. value investing works, but if everyone does it then people will undervalue intangibles. If everyone chooses to be honest, the first fibber will rule the world. If everyone is violent, the first non-violent group will waste fewer resources. Etc.

There is actually a lot of research into this area (game theory). And for a surprising number of situations the best and most stable strategy is somewhat random. EX: Throwing a fastball vs curve ball where the odds of hitting each depend on which was expected.

While it's clearly not a stable strategy if everyone did the same thing it's actually fairly hard to consistently beat a basket of randomly chosen stocks. Which is why I think rating stocks based on their relative value for your goals and then randomly picking a basket of them is probably the best and most stable strategy. (Excluding inside information.)

As long as you're smart enough to abstract out of your valuation system, you can survive. Buffett made his early money buying at a discount to assets: the company had a $5 million in the bank and their total market value was $3 million. Those opportunities mostly disappeared in the 1960's--conglomerates bought them out. But Buffett's meta-strategy of buying things at a discount still worked; it just worked better applied to businesses with a sustainable competitive advantage.

But if you're not exceptionally good at what you do, yes, some randomness would help. Though wouldn't the closest analogy to the fastball / curveball thing be to randomly alternate among asset-management paradigms, instead? Have some of your portfolio passively indexed, some invested for catastrophe, some in pie-in-the-sky growth, some in inflation-resistant consumer goods, etc.

tell that to the Goldman Sachs Alpha Fund

You may be able to profit from smaller opportunities than they can profitably exploit.