Almost every lesson on fake breakouts is a lesson in prediction: here are the tells, here is how to know in advance whether this one is real. That framing is the problem. A breakout and a fake breakout begin identically, by construction — and the traders who do well on them are not the ones who guessed right, they are the ones whose plan did not require guessing.

We tested the premise underneath that claim on 489 real breakouts and it came back split: the idea holds, and the casual version of it does not survive contact with the numbers.

Both outcomes start with the same event

When price closes beyond the edge of a range, the same sequence fires regardless of what happens next:

flowchart LR A([Close beyond the boundary]) --> B[Resting stop orders<br/>become market orders] B --> C[Breakout entry orders fill] C --> D[Momentum algorithms join<br/>the expanding move] D --> E([An impulse — in both outcomes])

Nothing in that chain knows whether the move will be accepted. Stops trigger because they were resting there, not because the breakout was correct. That is precisely why the impulse is common ground — and why it is the part of the move that can be monetised without a forecast.

The question is whether the impulse on a failed breakout is big enough to be worth anything. That is measurable, so we measured it.

The test: does a failed breakout still pay?

489 breakout attempts on SPY 1-minute bars across 490 sessions, taken only where a range had genuinely compressed first, between 09:45 and 14:30 ET. Every result is scaled by W — the width of the range that broke — so it compares across days.

Chart comparing the median best price reached after breakouts that held versus breakouts that failed, in range widths
Median best price reached after the breakout bar, in range widths. The failed group still climbs — a third as far, and it flattens almost immediately.
489
Attempts measured
75%
Failed within 30 min
0.43×
Failed impulse vs real, at +5m
49%
Failures reaching 0.25W in 10m
Median best price reached +3 min +5 min +10 min +30 min
Breakouts that held (n=124) 0.32W 0.43W 0.72W 1.29W
Breakouts that failed (n=365) 0.16W 0.19W 0.24W 0.41W

The premise holds. A failed breakout is not an instant reversal — it moves in your favour first. Half of them (49%) reach a quarter of the range width within ten minutes.

And the casual version dies. The failed impulse is roughly a third the size of a real one. Only 21% of failures reach half a range width, against 75% of the ones that held. And the median failure confirms itself in four minutes (p25 2, p75 9).

🚨 DANGER
That four-minute number is the whole discipline. If your plan is "enter the break, see how it goes, take something off if it works," you will still be deciding when the answer arrives. The harvest has to be **pre-placed and mechanical**, or on three attempts out of four there is nothing left to harvest.
How was "failed" defined, and does the definition change the answer?

Three ways, and the article reports the range rather than picking the flattering one. Any close back inside the range within 30 minutes: 83% of attempts. Two consecutive closes back inside — the one used above, because a single poke back inside is often a normal retest rather than a failure: 75%. Still inside the range at the 30-minute mark: 47%. The middle definition is the honest one for a scalper, and it is the strictest of the three that does not punish a healthy retest. The MFE gap between the groups is similar under all three.

What the numbers license: harvest, protect, participate

If the impulse is real but small and brief, and the outcome is unknowable at entry, then the correct response is not a better filter. It is a position structure that gets paid by the part you know exists.

flowchart TD A([Enter on the break]) --> B["HARVEST — sell the first tranche<br/>into the impulse, before the outcome is known"] B --> C["PROTECT — sell the second into extension,<br/>taking the campaign out of danger"] C --> D{"Does price accept<br/>outside the range?"} D -- "Retest holds" --> E([PARTICIPATE — the remainder rides]) D -- "Closes back inside" --> F([Exit the remainder])

The logic is uncomfortable for anyone trained to let winners run: you are selling into strength before you know whether the strength meant anything. That is the point. With a 75% failure rate, the impulse is the only part of the distribution that shows up reliably — so it is the part you convert into money, and the remainder becomes a free option on the 25% case.

Set the campaign and compare it against holding everything:

⛶ Full Screen

💡 TIP
This is also the honest argument for trading more than one contract. It is not about size. **With a single contract you can harvest or you can participate, never both** — and the whole method depends on doing both.

Two confusions that cause most fake-breakout losses

The draft this article came from ends up at a state machine, and the reason is that most losses are not bad analysis — they are actions taken at a state that did not permit them.

Armed is not entry. A compressed range with a defined boundary and clean room above it is a reason to have an order ready. Traders routinely convert it into a position before anything breaks, then defend a thesis the market never confirmed.

Impulse is not validity. The burst after the break feels like proof. It is not — it is the mechanical consequence of resting orders being triggered, and the data above shows it happens in both outcomes. Reading it as confirmation is what turns a harvest into a hold.

⛶ Full Screen

⚠️ WARNING
Compression deserves the same caution. It tells you where to look, not what will happen: in a separate test we found compression does **not** measurably predict expansion — a day-clustered paired difference of +0.0055 ATR, interval [−0.0071, +0.0182], n = 236 sessions. Use it to choose the level you watch, never as a forecast of the move that follows.

Build the boundary as a zone, from the wicks

A breakout can only be defined against a level, and the level is usually drawn wrong.

  • Do not drop the wicks. The wick extremes are where the resting orders actually sat. That is the price a breakout has to genuinely overcome, and ignoring it produces "breakouts" that never cleared anything.
  • Do not use one line either. Cluster the highs and the lows into a zone. Price interacts with an area of interest, not a pixel — the same conclusion we reached in why price reverses at key levels.
  • Then use the close to classify. A wick through the zone that closes back inside is a liquidity sweep. A close outside is an attempt. The wick defines the level; the body decides what just happened.
What you see What it is What it licenses
Wick pierces the zone, closes back inside Liquidity sweep Watch. The level held and someone's stops just paid for it
Close outside the zone Breakout attempt Execute the plan you already wrote
Two consecutive closes back inside Failure Exit the remainder; optionally arm the reverse
Retest holds the zone from the other side Acceptance The runner earns the right to stay on

A breakout in the right direction can still be a bad trade

The last filter has nothing to do with whether the break is real. It is whether there is anywhere to go.

If the nearest opposing level — a prior high, an unfilled gap, a supply zone, session VWAP — sits half a range width above your entry, then a correct breakout still cannot pay you, because the move ends before your first target. The runway has to be measured before the trade, not admired after it.

must fit inside the runway
Harvest target
distance to the nearest opposing level
Runway
no trade, however good the break looks
If runway < target

That check, plus the entry sequencing in our scalp entry checklist, removes more losing breakouts than any pattern-recognition improvement will.

The mistakes this framework is designed to prevent

What most traders doWhat the numbers support
On the breakenter and wait to see if it is realexecute a plan written before the break
On the impulseread it as confirmation, hold everythingharvest — it is the part both outcomes share
On the first pullbackmove the stop to breakeven, get tapped outlet structure, not a round number, decide
On failuregive it room, hope for a second attemptexit the remainder inside about four minutes
Sizingfrom confidence in the setupfrom risk, since 75% of attempts fail

The takeaway

  1. Stop trying to classify the breakout at entry. It is not knowable then, and the plan should not require it.
  2. Harvest the impulse. Both outcomes produce one; failures reach a quarter of the range width about half the time.
  3. Do it fast. The median failure confirms in four minutes, and the failed impulse is a third the size of a real one.
  4. Let the remainder answer the question. Acceptance outside the range is what earns a runner — not the impulse, and not your conviction.
  5. Size from the 75%. Three attempts in four close back inside. A plan that only works on the fourth is not a plan.

The uncomfortable truth in the data is that you were never going to tell them apart in advance. The good news is that you do not have to, provided you take something while the market is still deciding.

The one line to keep
Break wins the level. Retest defends it. The impulse in between belongs to whoever took money from it before the answer arrived.
ℹ️ INFO
**Method.** SPY 1-minute regular-hours bars, 490 sessions. A breakout attempt is the first bar between 09:45 and 14:30 ET that closes beyond a 20-bar range which had itself compressed to under 60% of the preceding window — 489 attempts, one per session at most. Outcomes are measured over the following 30 minutes and scaled by the broken range width W (median 1.00 point, about 0.15% of price). Failure is defined as two consecutive closes back inside the range; the alternative definitions are reported in the collapsible above. These are medians of an underlying-only study: no spread, no commission, and no option overlay, all of which move real results against you. One instrument, one timeframe.