Breakout Trend-Bar

does breakout context rescue the coin flip? — a pre-registered null

The trend-day study (H14) left a sharp loose end: after a daily trend bar the next session reliably tags a new extreme (early follow-through ~75%), but the next closedirection is a coin flip (~53%). H14 treated every trend bar as one thing. Al Brooks’ central teaching says that is the wrong unit — a trend bar is only an “always-in” breakout when it breaks out of the range: “a breakout that has follow-through becomes a trend; a trend bar inside a trading range gets reversed.” So we pre-registered the obvious rescue: split trend bars into those that close through the prior 20-day extreme (real breakouts) and those that stay inside it, and check whether the close coin flip finally breaks. Across 766 breakout trend bars and 2,011 in-range trend bars on 15 liquid US names (Jan 2022 – May 2026), it does not. The early poke is real; the close stays a coin flip either way. We publish the null because it tells you where the edge isn’t.

Breakout early follow-through
75.6%
tags a new extreme next session · CI95 [72.5%, 78.6%]

What survives: a breakout trend bar pokes past its extreme next day ~3 of 4 times, symmetric bull (75.9%) and bear (75.2%).

Breakout close stays in-trend
48.7%
CI95 [45.2%, 52.1%] — straddles 50%

The central claim’s null: the next close finishes in the breakout’s direction no better than a coin — if anything slightly worse.

vs. in-range trend bar
50.5%
context diff −1.8% · CI95 [−5.8%, +2.4%]

A trend bar stuck inside the range closes in-trend just as often. Breakout context moves the close needle by nothing.

What breakout context changes — and what it doesn’t

Early follow-through (tags a new extreme)

Breakout trend bar
75.6%
n=766 · CI95 [72.5%, 78.6%]
In-range trend bar
70.6%
n=2,011 · CI95 [68.5%, 72.5%]

Both well above the 50% you’d guess. Breakouts edge in-range bars (75.6% vs 70.6%), but BOTH have strong early follow-through. The next-day poke is a property of trend bars as a class, not of breaking out.

Close direction (finishes in-trend)

Breakout trend bar
48.7%
n=766 · CI95 [45.2%, 52.1%]
In-range trend bar
50.5%
n=2,011 · CI95 [48.2%, 52.6%]
Coin flip
50.0%

All three bars sit on top of each other at 50%. The breakout CI ([45.2%, 52.1%]) and the in-range CI ([48.2%, 52.6%]) both contain 50%, and the difference between them straddles zero. Breakout context tells you nothing about the next close.

15 tickers · 2022–2026 · n=5,000 bootstrap, seed 17. The split that should have broken the coin flip — Brooks’ own breakout-vs-range distinction — leaves it intact.

Per-ticker detail

Sorted by breakout count. Read the two right-hand close columns against 50%: they scatter from 36.4% (JPM) to 61.3% (AMD) on both sides with no pattern — textbook noise. The breakout-continuation column, by contrast, is high and steady (~75%) name by name: that is the part that is actually real.

TickerN breakoutBreakout follow-throughBreakout close in-trendIn-range close in-trend
AMD6279.0%61.3%44.9%
TSLA6075.0%48.3%47.4%
XLK5879.3%50.0%49.0%
XLF5474.1%48.1%49.2%
AVGO5466.7%40.7%49.3%
SPY5284.6%50.0%52.9%
MSFT5164.7%41.2%48.2%
IWM5074.0%58.0%51.1%
NVDA4973.5%51.0%50.0%
QQQ4776.6%51.1%50.0%
META4783.0%42.6%52.1%
GOOG4780.9%53.2%48.8%
AAPL4673.9%52.2%62.1%
AMZN4577.8%42.2%48.4%
JPM4470.5%36.4%54.1%

What this means for trading

  1. Don’t hold a breakout-day position overnight for the close. A daily trend bar that closes through a 20-day high feelslike the start of a directional run — that’s the always-in intuition. But the next session closes in your favor only 48.7% of the time, a coin flip. The breakout earns you the early poke past the extreme, not a follow-through close.
  2. Trade the early extension, not the trend. What does survive is the 75.6% rate that the next session tags a new extreme — intraday, early. That is a scalp/target against the prior extreme, not a swing thesis. The edge has a short shelf life and decays into the close.
  3. Breakout vs. in-range is not the dividing line you think it is.Brooks’ distinction is real on the chart, but at the daily close-to-close scale it does not separate a directional bar from a reverting one. Both close in-trend ~50%. If you filter trend-bar trades by “did it break out of the range,” you are filtering on a variable that carries no next-close information here.
  4. A null protects your capital.“A real breakout has follow-through” is one of the most intuitive lines in price action, and it is exactly the kind of belief that needs a number before you size on it. The number says: the early poke is real, the close is a coin flip — context and all.

Pre-registered verdicts

H16(a) — breakout follow-through > 60% both directions: PASSED. Bull 75.9%, bear 75.2% (overall 75.6%, CI95 [72.5%, 78.6%]). Breaking out does not kill the early next-day poke — it stays strong and symmetric.
H16(b) — context rescues direction (breakout close-cont > in-range, CI > 0): FAILED. Breakout 48.7% vs in-range 50.5%, difference −1.8% (CI95 [−5.8%, +2.4%]) — straddles zero, and points the wrong way. The central claim does not hold: breakout context does not make the next close predictable.
H16(c) — in-range trend bars are the coin flip (CI straddles baseline): PASSED (null confirmed). In-range close-continuation 50.5% vs 50.0% baseline, diff +0.5% (CI95 [−2.4%, +3.3%]). The coin flip H14 found is present in the in-range cohort — but, per (b), it is also present in the breakouts. It is not concentrated; it is everywhere.

The honest reading: (a) and (c) pass, (b) — the one that mattered — fails. Brooks’ breakout-vs-range distinction sharpens a chart read, but it does not buy a daily directional edge. The trend bar’s value is the early extension; the close is a coin flip with or without the breakout.

Honest caveats

  • This is a daily close-to-closetest. The breakout’s directional edge, if it exists, may live on the intraday path — a same-day continuation that is tradeable in the morning but mean-reverts by the close. A null on the close says nothing about an intraday-managed entry.
  • “Breakout” is defined as close beyond the prior 20-day extreme. A different lookback (10-day, 50-day), or requiring the breakout to clear the level by a minimum ATR margin, would isolate different events. Changing the rule after seeing this result would be the post-hoc tuning the pre-registration exists to prevent — that belongs in a new, separately pre-registered study.
  • Horizon is one session. A breakout’s directional follow-through could play out over 2–5 days even if day +1’s close is a coin flip. A multi-day holding-period sweep is a natural, honest extension.
  • Universe is the same 15 mega/large-cap US names as the prior studies — highly liquid, in a broadly bullish 2022–2026 window (note the ~50% directional baseline once bull and bear are folded together). Breakout trend bars in choppier regimes or small-caps may behave differently.

How this fits the other studies

  • Trend-Day Follow-Through: This is the direct extension of H14. H14 found early follow-through real (~75%) and the next close a coin flip across all trend bars. H16 asked whether breakout context is the hidden variable that splits the two — and finds it is not. The coin flip is robust to the most natural rescue.
  • Failed Breakout Fade: The mirror null. There, fading a breakout that fails is a coin flip; here, riding a breakout that holdsis a coin flip on the close. Together they bracket the same lesson: at the daily scale, the breakout’s success or failure carries no next-close directional edge by itself — the edge, if any, is in the intraday confirmation.
  • Inside Day / “ii”: Those studies measure compression flagging magnitude (expansion), never direction. H16 is the matching result on the expansion side: a trend-bar breakout flags an early move, never its close direction. Structure predicts how much, not which way — over and over.

Methodology

766 breakout trend bars + 2,011 in-range trend bars on 15 names (2022-01-03 – 2026-05-30) via Massive/Polygon daily bars. Trend bar (identical to H14): body ≥ 50% of range, close in the top (bull) / bottom (bear) quartile of the range, and range ≥ Wilder ATR(14) measured strictly before today. Breakout: a bull trend bar whose close exceeds the max high of the prior 20 bars (mirror for bear lows); the 20-bar window excludes today, so the breakout is a property the bar prints, not lookahead. In-range: a trend bar whose close stays inside the prior 20-day range. Early follow-through: next.high > trend.high (bull) / next.low < trend.low (bear). Close continuation: next.close finishes in the trend bar’s direction. Direction-folded baseline: directional close-rate across all qualifying day-pairs. Bootstrap CI (n=5,000, seed=17). Engine: scripts/ml/backtest_breakout_trend_bar.py. Run 2026-06-25.

Related: Trend-Day Follow-Through · Failed Breakout Fade · Inside Day Expansion