I should also mention that academics rarely have access(*) to this data and therefore have to assume that the market figures out the right prices. The EMH is an extremely convenient assumption for academic research. It's the job of traders to assume it's it doesn't hold and then proceed to make it hold.
(*) Closing prices are universally available and practically free since decades. Intraday, not so much. Level2 data, not at all, and still super expensive.
As for numbers ... so mathematically, WLOG, you only need 50%+ correctness to make money. Below 50% no money can be made. How the monies are made depends on the distribution of the outcomes. Lets first suppose the distribution of outcomes is uniform, that is, you can't really tell whether one trade is better than another. In such a case, you can make money on a 50.1% correct model by spinning and hedging. That is, trading as rapidly as possible while ensuring you remain fully hedged at all times (or as much as possible). This is in the main how most algorithms work. They're just a few epsilon smarter than "random" (aka 50%) but they capitalize on that by trading a lot. A typical algo trades more in a day than humans do in a decade.
If you were a human, you can do the same, but you could also diversify (aka poor man's hedging). If you're winning on 60% of your positions and losing on 40%, you put equal amounts in each position. You're now beating the market (alpha) while cutting risk through diversification. This would be a very typical way to go about it.
Now, if you actually have a distribution of the outcomes, that is, you have more faith in some trades than others. Now you can start swinging for the fences. Normally mutual funds are constrained to diversify ... but if you look at so-called "conviction" buys from managers, you'll see they do better than random. If you were a private investor or representing private money (not public) you'd be smart or/and brave to concentrate on your conviction buys. Insofar your conviction is correct, then you can do much better.
Add: I should also say that those 60% would vary not necessarily by positions but possibly by years or even eras. There are many value-investors that have thrown in the towel because with CAPE10>30+, they've been wrong for years now. Conversely, momentum indexers have been right for almost a decade now and can seemingly do nothing wrong. Often you need to have not only the right talents but also have them at the right time.
)