Failed breakout at a level

It is 10:15 ET (22:15 ICT). ES trades above the level you marked before the open, and the chat says “breakout”. Eight minutes later price is back under the level, the probe has left a wick 2.5 points above it, and almost no volume traded above it. Was that a gift, a trap, or just noise? This lesson turns the moment into a written rule you can execute and score: what the trigger is, where you are wrong, where you take profit, and what the data says about whether the rule has an edge. Short answer for the last part: on NQ, it has not been shown to.

Key term

Failed breakout (phá vỡ thất bại): price probes beyond a mapped level, fails the acceptance checks you wrote in advance, and returns inside Definition. The failed breakout is the event; the fade (giao dịch ngược hướng phá vỡ) is the trade you may take against the probe once it has failed.

The anatomy of the fade

Every setup in this module is written in the same five parts: context → trigger → invalidation → target → R:R. A setup missing a part is a hunch with a chart.

• Context: a level you mapped before the open (M4.07): the prior day’s high or low, the overnight extreme, the initial-balance edge. The map tells you where to watch, not where price turns House rule. As the evidence below shows, mapped levels and confluence zones have not been respected more often than random prices, so a bold zone on your sheet is not a reason to size up.
• Trigger: the probe fails the acceptance checks of M4.01 on a window that matches the trade. This setup is a short-horizon trade, so it uses M4.01’s short window, written in advance: back inside within 15 minutes of the first trade beyond, volume beyond the level below the usual for that time of day, and a 5-minute close back inside House rule. These numbers are conventions that make the check scorable, not measured cut-offs.
• Entry: on the return inside, not on the wick. The wick is a probe in progress; the return inside is the verdict House rule.
• Invalidation: beyond the probe extreme, plus a small buffer. If price takes out the extreme and is accepted there, the idea is dead: exit, no debate. Placing the stop close is what keeps the risk small; it says nothing about how often the stop is hit.
• Target: measured from the map before entry: the nearest reference inside (VWAP, the initial-balance midpoint, the prior value-area edge). The opposite side of the range is not a default target: on NQ, failed breaks did not reach it more often than comparable moves (evidence below).
• R:R and break-even: the stop and target together imply a break-even win rate. Compute it with costs before the click (M6.08 covers stops and targets). If you cannot argue that the rule wins more often than that, the math is not “good”; it is unknown.

Because the verdict window is short, the trade is short too: if the target has not traded within about an hour, close the position at market House rule. A 15-minute verdict says nothing about the rest of the day.

Myth / hypothesis: “The failed breakout traps the breakout buyers, and their stops fuel the reversal.” Lore It is a good story and a useful picture of who is losing money, but stop locations cannot be observed in the data, and the test that comes closest (failed breaks reaching the other side, below) showed no edge on NQ. Treat the trapped-trader story as description, not as a reason the trade works.

Myth / hypothesis: “The better the level, the better the fade: at a confluence zone the failed breakout is the auction’s verdict.” Lore Our lab measured level families and confluence zones against random prices and found no difference (next section). Level quality, as usually scored, is not yet shown to change the outcome.

Why it is still worth knowing

• Shared vocabulary. “Probed the prior high, refused in eight minutes, back inside” is a precise description of the auction.
• Framing risk. The probe extreme gives you a nearby, objective invalidation. Even with no edge shown, a defined stop is how you keep each attempt small.
• A testable rule. Written checks, a fixed entry and a fixed exit can be scored on replay and in your journal (see the drill). That is the only honest route from “unproven” to “measured”.

What the data says

The setup’s popular promise is that a failed breakout reverses and runs back through the range. On NQ that promise has been tested once, in a public practitioner study, and was not supported.

So the honest status of this setup is: a rule-based fade with a pre-set stop, whose edge is unproven. You can still trade it as a small, defined-risk rule while you score it. You cannot call it high odds.

Worked example: the 6,024.50 wick

Synthetic session, ES (0.25-point ticks, $12.50 a tick, $50 a point); it shows the reading, not a signal.

• Map, written before the open. Resistance zone 6,018–6,022 (prior-day high and overnight high). Scenario: if a probe above 6,022 fails the short-window checks, fade it back to VWAP; no trade if it is accepted.
• 10:15 ET (22:15 ICT), probe. First trade above 6,022. Description: buyers testing the level; no verdict yet.
• 10:17 ET, probe high 6,024.50. Volume above the level is thin for that time of day.
• 10:23 ET, verdict. The 5-minute bar closes at 6,021.00, back inside, 8 minutes after the first trade above. Time ✗ (8 minutes, inside the 15-minute window), volume ✗ (below usual), close ✗ (back inside). Failed acceptance on all three checks.
• Plan. Short 1 ES at 6,021.00 on the return inside. Stop 6,025.00: one tick beyond the probe high plus a small buffer, 4 points = $200. Target VWAP 6,010.00: 11 points = $550. Time stop: flat by 11:25 ET if the target has not traded.
• Break-even. With $15 of round-turn costs, the trade nets +$535 on a win and −$215 on a loss, so it needs to win 215 ÷ 750 ≈ 29% of the time to break even. How often this fade reaches VWAP on ES or NQ is not yet measured, so you do not know whether your rule clears 29%.
• Outcome on this synthetic day. VWAP trades at 11:05 ET; +$535 after costs. One day is one outcome, not evidence.

What made this a plan rather than a gamble: the level was mapped before the open, the checks were written before the probe, and the stop, target and break-even were known before the click. Remove any one, and you have no way to score the result later. And note the trap this rule avoids: shorting every new high. Without a written check that failed, you are fading initiative.

The mirror and the failure

The long side (mirror). Below a mapped support, the same rule reads upside down: a probe under the level that fails the short-window checks and returns inside is a candidate long, with the stop beyond the probe low and the target at the nearest reference above. Flip the figure to see it. The evidence is the same: the public study’s test covered failed breaks of both the prior high and the prior low.

The failure. Same morning, different ending. After the short at 6,021.00, price rotates back above 6,022 at 10:40 ET, takes out 6,024.50, and the next two 5-minute bars close at 6,025.50 and 6,027.00 with volume building above. The stop at 6,025.00 fills: −$215 after costs. The read at 10:23 was correct; the auction changed its answer. A late acceptance is a normal outcome of this setup, not a broken rule, and it is the reason the stop exists. If the probe is now accepted on the longer window, the idea to watch is the opposite one, a breakout–retest (M5.07), which is itself only a lead, not an edge.

Prop-firm note. In a prop evaluation the stop must be a real order placed with the entry. Evaluation accounts usually fill in simulation, and simulated fills are not live fills; tuning very tight brackets or automatic breakeven moves to win on simulated fills is treated as prohibited conduct by firms such as Topstep. The Prop Firm course (PROP1.04) covers the rules; check your firm’s help center before you trade.

Common mistake: Entering on the wick instead of the return inside. The wick alone is a probe in progress; it can extend, build volume, and be accepted. Traders who short the wick’s tip are short into initiative with no verdict, and if it does fail, they entered before the evidence and get shaken out on any retest of the extreme.

Common mistake: Treating the setup’s popularity as an edge. “Failed breakouts reverse” is one of the most repeated lines in order-flow teaching; on NQ the tested version showed −1.1 percentage points against comparable moves. A setup that is easy to see and easy to name is not a base rate.

Common mistake: Sizing up at a bold confluence zone. Confluence did not change how often levels were respected in our lab’s placebo test. Keep size fixed by your risk rule, not by how important the level looks.

Drill: score twenty failed breakouts

1. Write your three short-window checks, your entry (return inside), stop (beyond the probe extreme plus one tick) and target (nearest reference inside) before you open a chart.
2. In replay, find 20 probes beyond a level you would have mapped before the open. Score each probe: failed, accepted, or undecided.
3. For every failed probe, record whether the target or the stop traded first, and the result in R after costs.
4. Compute your win rate and compare it with the break-even rate your stop and target imply.

What we don’t know yet

• How often the return-inside fade reaches its first target (VWAP or the nearest reference) on NQ, against a placebo entry at the same time: not yet measured.
• Whether waiting for the return inside beats entering on the wick, or not trading at all: not yet measured.
• Whether an order-flow trigger at the level (delta, absorption, a footprint pattern) changes the outcome: not yet measured as a rule; M4.09 shows how to write one so it can be scored. One related lead in the public study, delta divergence into a level’s first touch, is still graded Lore until a forward test.

Recap

• A failed breakout is a probe beyond a mapped level that fails written acceptance checks and returns inside; the fade is entered on the return inside, never on the wick.
• Write the plan before the probe: checks, entry, a stop beyond the probe extreme, a target from the map, a time stop, and the break-even win rate after costs.
• On NQ, failed breaks did not reach the other side more often than comparable moves, and mapped or confluence levels were respected no more than random prices: the edge is unproven.
• Late acceptance is a normal outcome; the stop handles it, and the opposite setup becomes the one to watch.

Quiz

Further reading

• Order Flow, Measured: all claims: the public practitioner study behind the first evidence box (claim AP-04).
• Auction Market Theory: failed auctions and acceptance in auction-theory terms.
• Our lab: 226 pre-registered tests on 721 sessions found no pre-market level family, and no confluence of families, respected more often than random prices at the same distance; on NQ and ES futures, prior-day levels were respected less.

Prerequisites: M4.01 Acceptance, M4.07 Daily, weekly & monthly level map, and EV3.01 on sample size. Next: M5.02 Absorption reversal, the same fade with an absorption print added, and what that addition does (and does not) change.

Educational content only, not financial advice. Futures trading involves substantial risk of loss. Synthetic examples are illustrations, not trade calls.