BUILD: COURSE 1 | LESSON 5
Chart patterns: continuation and reversal
Learning objectives
Identify the major continuation (flags, pennants, triangles) and reversal (head and shoulders, double tops/bottoms) patterns and the market logic behind each
Locate the trigger, stop and measured target for a pattern before entering — and size the trade accordingly
Calibrate expectations with realistic hit rates, including why most patterns fail and why failed patterns are information
Patterns are structure with nicknames
Everything in this lesson is Lesson 1 wearing a costume. A "double top" is a failure to make a higher high followed by a broken swing low. A "bull flag" is a trend pausing in a tidy counter-slope channel. The nicknames survive because they compress a lot of structure into one word — and because enough traders watch the same shapes that reactions around them are partly self-fulfilling.
Two rules before the catalogue:
Patterns only mean something in context. A head and shoulders needs a prior uptrend to reverse; a flag needs a strong impulse to continue. The same squiggle in the middle of a random range is noise with a name.
A pattern is not a trade until it completes. Every pattern below has a trigger level (usually a neckline or boundary break). Before the trigger, you have a hypothesis. Anticipating completion — shorting because the right shoulder "looks done" — is how traders end up positioned against a trend that simply resumes.
And the honest framing for the whole lesson: the most systematic public study of chart patterns (Bulkowski's work, tens of thousands of hand-audited cases in stocks) finds that even the best patterns fail their basic follow-through test roughly 25–40% of the time, and typical patterns hover near coin-flip usefulness once you account for slippage and where stops must go. Patterns don't print money. They give you locations where risk can be defined tightly — the edge, as ever, comes from the reward-to-risk geometry, not the shape.
Continuation patterns: the trend takes a breath
Flags and pennants. After a sharp impulse (the "pole"), price drifts in a narrow counter-trend channel (flag) or converging coil (pennant) for a handful of bars. Logic: early longs take profit, but pullback buyers absorb the selling without giving back much ground — a sign demand is still in charge.
Worked example: EURUSD rallies impulsively from 1.0700 to 1.0780 (an 80-pip pole), then drifts down for four days in a tight channel between 1.0765 and 1.0745. Trigger: a break above the flag's upper edge, say 1.0768. Stop: below the flag low at ~1.0740 (28 pips of risk). The measured move convention projects the pole from the breakout: 1.0768 + 80 ≈ 1.0848 target — roughly 2.8R if it gets there. That geometry, not the flag's prettiness, is the trade's justification. Caveats: flags that drift too long (10+ bars) or retrace more than about half the pole lose their "brief pause" logic; and in ranging regimes "flags" are just wobbles.
Triangles. Price coils between converging boundaries: ascending (flat top, rising lows — buyers pressing against a known seller), descending (flat bottom, falling highs), symmetrical (both converging — genuine indecision). Ascending triangles lean bullish and descending lean bearish, but the lean is mild; symmetrical triangles genuinely break either way, which is why the only disciplined play is to trade the break, not the guess. Trigger: close beyond the boundary. Stop: inside the triangle (mid-coil or beyond the opposite boundary, ATR-buffered per Lesson 4). Target: the height of the triangle's widest part projected from the break. The classic failure mode is the fakeout — a poke through the boundary that immediately reverses; triangles late in their apex, where the coil has gone quiet and narrow, fake out the most. Waiting for a close beyond the level, or a break-and-retest, trades a worse entry price for fewer traps. Neither approach eliminates them.
Reversal patterns: the trend changes its mind
Head and shoulders (H&S). The most storied reversal: in an uptrend, a peak (left shoulder), a higher peak (head), then a lower peak (right shoulder), with the intervening lows joined by a neckline. Read structurally it's simply: last higher high (head) → lower high (right shoulder) → and the neckline break is the lower low that confirms the reversal, exactly Lesson 1's two-stage sequence. The inverse H&S mirrors this at bottoms.
Worked example: gold tops. Left shoulder 2,410, head 2,442, right shoulder 2,415; neckline across the swing lows near 2,378. Trigger: daily close below 2,378. Stop: above the right shoulder, 2,420 (≈42 dollars of risk). Measured target: head-to-neckline height (2,442 − 2,378 = 64) projected down → ≈2,314, about 1.5R. Note what the measurement just told you: even textbook H&S geometry often yields mediocre R:R because the stop must live above a distant shoulder — one reason many traders wait for the frequent retest of the broken neckline (short the pullback to ~2,378, stop above ~2,400) to buy a better ratio at the cost of sometimes missing the move entirely. That trade-off has no free answer.
Double tops and bottoms. Two peaks in the same zone (they will not be to-the-pip equal — treat the second peak anywhere within a few pips-to-fractions-of-ATR of the first as "the zone") separated by a trough; trigger is the break of that trough. A double top at 1.2780/1.2775 on GBPUSD with the intervening low at 1.2690 triggers below 1.2690, stop above 1.2785, target 1.2600 (the pattern's height projected). Beware the most common real-time error: calling every pullback-and-return a "double top". Until the middle trough breaks, what you're looking at is a normal uptrend making a pause.
Hit rates, failed patterns, and how professionals actually use this
Time for the numbers nobody puts in the brochure. Across the large pattern-performance literature: measured-move targets are reached in very roughly 50–70% of completed cases depending on pattern and regime; a substantial minority of breakouts — commonly cited around 10–25% for the better patterns, worse for sloppy ones — fail outright and return through the pattern; and results measured in stocks with trends and volume data translate imperfectly to 24-hour FX, where volume is fragmented and whipsaws are common. Add spread and slippage and the honest summary is: patterns are weak predictors and decent trade-organisers.
So use them the way professionals do:
- As risk maps, not prophecies. The pattern's value is that it hands you a trigger, an invalidation point and a target before entry — everything Course P2's expectancy maths needs. If the geometry doesn't offer at least ~1.5–2R after realistic stops, the pattern being "textbook" is irrelevant.
- With the regime filter on. Continuation patterns with the higher-timeframe trend materially outperform counter-trend reversal hunting. The reversal patterns you should respect most are the ones forming at levels that already mattered (Lesson 1 structure, prior daily highs/lows).
- Failed patterns are signals too. A breakout that collapses back through a triangle traps every breakout trader; their stops fuel the move the other way. Some traders trade only these failures. At minimum: when your pattern fails, take the stop and note that the market just told you who's really in control.
- Beware pattern pareidolia. Squint at any chart and you'll find heads, shoulders and flags everywhere — mostly in hindsight. If you need to argue for the pattern's existence, it doesn't exist for the crowd, and the crowd is what makes patterns work at all.
Key takeaways
Every chart pattern is swing structure with a nickname; it needs a prior trend, and it isn't a trade until the trigger level breaks.
Continuation patterns (flags, pennants, triangles) mark pauses; triggers are boundary breaks, stops go inside the pattern, and measured moves project the pole/height.
Reversal patterns (H&S, double tops/bottoms) are Lesson 1's lower-high-then-broken-low sequence; the neckline/trough break is the confirmation, and retests often offer better R:R at the cost of missed trades.
Realistic performance: targets hit maybe 50–70% of the time, outright failures 10–25%+ — patterns organise trades, they don't predict.
The edge lives in the geometry (trigger, stop, target ≥ ~1.5–2R) and the regime filter, never in the shape itself; failed patterns are tradeable information.