A theory that explains everything afterwards predicts nothing beforehand

Two things that both feel like understanding

An explanation accounts for something that has already happened. A prediction commits to something that has not. Both produce the sensation of comprehending a system, and only one of them can be wrong in a way you will find out about.

This asymmetry is why fields that traffic in explanation can accumulate confident accounts for decades without ever discovering whether they’re correct. Every event that occurs is absorbed and narrated; no event is ever left unexplained; and the absence of failures is read as success rather than as the absence of any test.

Flexibility is the property that gives it away

The diagnostic question about any account isn’t whether it explains what happened but whether it could have explained the opposite. If a rise and a fall would both have been comfortably accommodated, then the account contains no information about which one to expect, whatever it contains about why the actual one occurred.

This is the practical version of falsifiability, and it does not require any philosophy of science to apply. Ask what pattern of events the theory rules out. If nothing is ruled out, the theory is a vocabulary for describing outcomes rather than a claim about them — which can still be useful, since vocabularies organise thinking, but it’s a different product and should not be sold as the other one.

Why after-the-fact explanation is so easy

Because the outcome supplies the shape. Once you know how something ended, the events leading to it reorganise themselves into a sequence pointing that way; the signals that pointed elsewhere become noise and the ones that pointed here become causes. The reorganisation is automatic and invisible, which is why post-event accounts feel discovered rather than constructed.

The same mechanism operates on the person doing the explaining, so sincerity is no protection. A commentator producing an account of yesterday genuinely believes they’re analysing rather than confabulating, and the account may well be correct. The problem is that a correct account and a plausible one are indistinguishable in this format, because the format contains no test.

What a prediction costs the person making it

A prediction has to name an event, a time and a confidence, and it has to be recorded somewhere the person can’t revise. Every one of those is expensive: the naming forces precision that vague understanding cannot supply, the timing removes the option of being eventually right, and the record removes the option of remembering it differently.

This expense explains the market. Explanation is abundant because it is cheap and safe, prediction is scarce because it is costly and risky, and the audience for both largely cannot tell the difference. Anyone willing to pay the cost is offering something rarer than the confident narration that fills the same slot.

The middle case that deserves respect

Not every valuable account is predictive, and demanding predictions from everything is its own error. Historical explanation, diagnosis of a particular failure, and interpretation of a complicated situation are legitimate activities that do not produce forecasts, and holding them to a forecasting standard would discard a great deal of real knowledge.

The requirement in those cases is different but not absent. A good after-the-fact account says what evidence would have contradicted it, distinguishes the parts it is confident about from the parts it is reconstructing, and ideally predicts something incidental — a record that should exist, a detail that should be found, a similar case that should show the same pattern. Those are testable consequences of a claim about the past, and an account offering none of them is difficult to distinguish from a story.

Applying this to your own understanding

The check is available privately and is unpleasant enough that people avoid it. Before an outcome you are confident about, write what you expect and how sure you are. Afterwards, compare. If you have never done this in a domain where you feel knowledgeable, then your sense of understanding that domain has never been tested, and the feeling is coming from the coherence of your account rather than from any record of it being right.

The uncomfortable finding, when people do try this, is usually not that they are badly wrong. It is that they are much less certain than they felt, and that the events they would have explained fluently afterwards were not ones they would have called in advance. That gap is where the difference between explanation and prediction actually lives.