A divergence on its own says nothing about what to do: it has no moment, no level and no deadline. It reports one thing — the current move's impulse is weaker than the previous one by the chosen indicator's measure. Everything that turns this into an action is added from outside.
Below is what people add, and which questions remain.
What a mismatch lacks to be a signal
A signal needs three things, and a divergence has none of them.
A moment. A mismatch exists as long as both extremes exist — it is a state, not an event. It can persist for weeks and renew at every new peak.
A level. A divergence names no price: not where to act, and not where to conclude you were wrong.
A deadline. "The impulse weakened" implies nothing about when or how that ends.
What people typically add
Three approaches are the most common, and all three are about waiting for an event.
Waiting for confirmation from structure. The mismatch is treated as background, and the decision comes from a break of a level or an extreme. The moment and the level then come from structure, while the divergence only selects which breaks to consider.
Waiting for a reversal bar or an exit from a zone. The event becomes a specific candle or price leaving a pre-marked area.
Placing the mismatch alongside other observations and deciding on the whole, without treating it as a standalone argument.
What they share: the event does not come from the divergence. It stays a filter rather than a trigger.
Why "entering on a divergence alone" ends badly
Because a mismatch is cancelled by nothing except a new extreme. A position opened on it has no natural point at which the idea is admitted wrong — and therefore no way to bound the loss by the logic of the markup itself.
The second reason: mismatches are plentiful. Any instrument holds dozens of them, and most are followed by nothing notable.
What we will not advise
There are no specific entry, exit or position-sizing rules here, and there will not be. Markup describes what has already happened on a chart; what to do about it is a decision that depends on goals, horizons and acceptable risk, and it does not follow from the shape of a curve.
A common mistake
Concluding that because a divergence "did not work", it was false. A mismatch is not a forecast, so it cannot "work". It was and remains a correct description of the past; the expectation of a reversal was added by you.