Eugene Fama's formulation in the 1960s: asset prices incorporate the information available about them. If everyone can see that a company will do well, that is already in the price, and buying it earns no excess return.

Three strengths are usually distinguished — prices reflect past prices, or all public information, or all information including private. The middle version is the one that is argued about seriously.

Why the mechanism generalizes

The reasoning is not really about finance. It is about competition for a known opportunity.

If a way of making money is visible, legal and does not require anything scarce, people will do it until the return falls to the cost of doing it. Persistent easy returns imply something preventing entry.

That makes the hypothesis a diagnostic question rather than a doctrine: when you find an apparently excellent opportunity, ask why it is still available. There are good answers, and the answer is the actual thing you are relying on.

The good answers

It requires something scarce. Capability, a relationship, a license, accumulated knowledge. This is a real edge and it is what positioning and career capital are for.

It is unattractive. Boring, unglamorous, low-status work is systematically underserved, because the people who could do it would rather do something else. This is one of the most reliable sources of return available to a small business.

It is too small. Below the threshold larger competitors can profitably serve. Most of what a good small firm does lives here.

It is genuinely new. The window before competition arrives, which is real and short, and is the whole of creative destruction from the entrant's side.

You are early to information that is not yet widespread. Real, and rarer than people think.

The bad answer

Everybody else has missed it. Occasionally true and usually a sign that the reason it is available has not been found yet. The disciplined version of the question is not "is this a good opportunity" but "what do I have that the people who did not take this lack?"

Where markets are visibly not efficient

The hypothesis is an approximation and the exceptions are informative. Prices move more than information seems to warrant, momentum persists longer than it should, and bubbles occur and are visible in hindsight. The theoretical objection is neat: if prices already reflected everything, nobody would be paid to research them, and without that research the prices would not reflect anything.

So markets are efficient roughly to the degree that people are paid to make them so. Which is why the concept transfers usefully: an opportunity in a market nobody is studying is far likelier to be real, and a market with many well-resourced participants competing on the same public information is one where the easy returns have already been taken.