Confirmation and invalidation
Two questions to answer before entering: what would show this is working, and what would prove it wrong.
11 min read · intermediate · lesson 5 of 5
What this covers
- State a confirmation condition before entering
- Locate the invalidation level for a position
- Distinguish price invalidation from time invalidation
- Explain why a correct analysis can still produce a loss
- Keep the sizing decision separate from the analysis
The two questions
- Confirmation
- The observable event that has to occur before the read is acted on. A close beyond a level, a structure break in the expected direction, a rejection at a zone.
- Invalidation
- The price at which the read is wrong. Not the point of maximum tolerable loss, which is a separate and smaller number.
- Time invalidation
- The point at which a read has had long enough to work and has not. A thesis that required an immediate reaction and got a week of drift has been answered, even without the price level being reached.
What confirmation is for
A read describes a condition. Confirmation is the evidence that the market has begun to act on it, and it exists to filter reads that are correct in principle and early in practice.
The cost is real. Waiting for confirmation means a worse entry price than anticipating the move, and some moves run without giving one. What it buys is the removal of positions taken on a view that the market never validated, which is the larger category.
Define the confirmation before entering, in terms specific enough that it either happened or it did not. A condition that can be argued after the fact is not a condition.
What counts as confirmation
The useful ones share a property: they are visible, binary, and cannot be reinterpreted after the fact.
A close beyond a level qualifies. Either the period closed past it or it did not. A structure break qualifies for the same reason. A decisive reaction at a zone qualifies, provided decisive was defined in advance rather than judged afterwards.
What does not qualify is anything requiring a judgement call at the moment of acting. The point of setting the condition beforehand is to move the decision out of the moment when the position is most tempting, and a condition that still needs interpreting has not moved it anywhere.
Invalidation is a level, not an amount
Invalidation comes from the analysis. If the case rests on a zone holding, invalidation is a close beyond that zone. That level exists regardless of position size and regardless of the account behind it.
The sequence is: find the invalidation level, measure the distance from entry, then size the position so that distance costs an acceptable amount. Running it in the other direction produces a stop placed at whatever distance the desired size allows, which puts the exit somewhere the analysis says nothing about.
A position with no identifiable invalidation level is a position with no thesis. That is the signal to skip it rather than to widen the stop.
A read can expire without being wrong
Most invalidation is discussed in price, and the more common outcome is neither target nor stop. The position simply sits.
A read usually contains an implicit expectation about timing. A break expected to run should run reasonably soon; the participants who drove it are still there. When a week passes and nothing has extended, the conditions that produced the read have changed even though the invalidation level was never reached.
Deciding that horizon in advance turns an open-ended position into a closed question. Without it, the default outcome is holding until the stop, which converts every read that quietly stopped being true into a full loss.
A correct read can still lose
Analysis describes the balance of current evidence. Evidence favouring one direction does not mean price moves that way next, and a single outcome carries no information about whether the process producing it was sound.
This is the reason risk controls sit outside the analysis rather than inside it. The analysis decides what to do and where it would be wrong. Position sizing decides what a wrong answer costs. The second job has to work when the first one fails, which is most reliably arranged by keeping them separate.
Judge a process across a sample of outcomes. A run of losses on well-executed positions is an ordinary property of a sound process, and a run of gains on poorly-sized ones proves nothing.
Both questions answered
IllustrativeLevels below are illustrative.
Read: a market in an uptrend has pulled back into a zone at 1.0940 to 1.0960 that previously capped price.
Confirmation: a close back above 1.0960 after trading into the zone. Without it there is no entry, however good the zone looks.
Invalidation: a close below 1.0910, beneath the zone and beneath the last higher low. At that point the pullback is a structure break and the read is wrong.
Time invalidation: five sessions. If the zone has neither produced the close above nor failed by then, the setup is stood down rather than carried.
Sizing: entry near 1.0965 and invalidation at 1.0910 gives 55 pips of risk. Position size follows from that distance and the risk per trade already fixed.
All four answers exist before the position does. None of them move after it opens.
Key takeaways
- State confirmation before entering, in terms that either happened or did not.
- Useful confirmation is visible and binary. Anything needing a judgement call in the moment is not.
- Invalidation is a level from the analysis, not an amount from the account.
- A read can expire on time without the price level being reached. Decide that horizon in advance.
- No identifiable invalidation level means no thesis. Skip it.
- A correct read can lose. Keep sizing separate from analysis so one works when the other fails.
Common mistakes
- Moving the invalidation level after entry because price approached it.
- Defining confirmation loosely enough that any candle qualifies.
- Holding a stalled position to the stop rather than standing it down when the thesis expired.
- Judging a process on one outcome rather than on a sample.
Knowledge check
OptionalRelated
- Support, resistance and decision zonesMarket Intelligence
- Positioning and market participationMarket Intelligence