Continuation patterns are brief consolidations that interrupt an existing trend and are expected — by convention — to resolve in the direction of that trend. The two families covered here are (1) flags and pennants: short pauses (typically one to twelve weeks per StockCharts ChartSchool) following a near-vertical "flagpole" advance or decline; and (2) triangles: converging price ranges that take one of three shapes — symmetrical, ascending, or descending. The core tension is that the "continuation" label is partly folklore: measured base rates show that breakout direction is far less predictable than the names imply, and a "valid" pattern is only as reliable as the strictness with which you define it. Treat the lore as a prior, not a verdict — the measured numbers below are the discipline.

How they're formed

Flag — a sharp, high-volume move (the flagpole) followed by a small rectangular consolidation that drifts counter to the trend: an up-flagpole produces a flag that slopes gently down; a down-flagpole, a flag that slopes up (StockCharts). The trigger is a close beyond the flag's bounding trendline in the flagpole's direction.

Pennant — identical setup to a flag (flagpole + brief pause), but the consolidation is a small symmetrical triangle converging to a point rather than a parallel channel. Distinguishing a flag from a pennant is cosmetic; they're measured and traded the same way.

Triangles — converging trendlines over a longer base (weeks to months):

• Symmetrical: a falling upper line and rising lower line (lower highs and higher lows) — pure compression with no built-in directional bias.
• Ascending: a flat horizontal resistance line with a rising lower line — read as buyers pressing into a fixed supply ceiling.
• Descending: a flat horizontal support line with a falling upper line — read as sellers pressing into a fixed demand floor.

Volume is part of the textbook definition: heavy on the flagpole, contracting through the consolidation/triangle, then expanding on the breakout (StockCharts, Bulkowski). A breakout on dry volume is the most common false signal.

The breakout trigger and the measured-move target

The entry trigger is a decisive close beyond the pattern boundary in the breakout direction. The conventional price objective is the measured move (a.k.a. the measure rule):

• Flags/pennants — measure the flagpole height (start of the swing to the consolidation) and project it from the breakout point. Bulkowski computes the height from the swing start (A) to end (B), then adds it to the flag bottom (up-break) or subtracts from the top (down-break) (thepatternsite.com/flags.html, /pennants.html).
• Triangles — measure the height of the triangle at its widest (the base) and project it from the breakout point (StockCharts; Bulkowski measures peak-to-valley and adds/subtracts from the breakout price).

A practical refinement Bulkowski stresses: the full measured move is hit far less than half the time (see below), so the raw target is an optimistic ceiling, not an expectation.

How they're used in practice

The canonical, style-agnostic application: (1) confirm a prior trend and a clean consolidation with contracting volume; (2) wait for a breakout close beyond the boundary, ideally on a volume expansion; (3) project the measured move as a reference target while scaling out before the full target given the low completion rates; (4) place the invalidation stop just inside the pattern (below the flag/triangle for longs). Many traders treat the opposite-direction break as a tradeable signal in its own right rather than just "noise," precisely because triangles break against the lore so often. Exact entry, stop, and hold-period mechanics for swing trades are deferred to the Swing Trading branch — this node covers structure and base rates only.

Standing & evidence — folklore vs. measured

This is where these patterns most need honesty. Two specific corrections matter.

(a) Triangles frequently break against the lore. Per Bulkowski's measured stats (bulk U.S. stocks, thepatternsite.com):

The headline: the descending triangle — the "bearish" one — actually broke upward 53% of the time in Bulkowski's sample, and the ascending triangle breaks up only ~63%, not the near-certainty the bullish lore implies. Symmetrical triangles are not a coin-flip (Bulkowski measured ~60/40 up), but they carry no reliable directional edge and rank near the bottom of his 39 patterns for post-breakout performance. Bulkowski also notes triangle performance has decayed over the decades (descending-triangle performance "almost [halved] since the 1990s"). The honest read: a triangle tells you compression is happening; it does not reliably tell you which way it resolves — trade the actual break, not the shape's name.

(b) Flag/pennant reliability is overstated and definition-sensitive. Bulkowski declines to assign flags and pennants a performance rank at all, because their move is measured against the short-term swing rather than to an ultimate high/low like other patterns — so the impressive-sounding rankings simply don't exist for them. His measured figures are modest: flags break up ~60% of the time with a 44% break-even failure rate, ~9% average rise, and only 46% hitting the measured-move target; pennants are weaker still — ~57% up, a 54% break-even failure rate, ~7% average rise, and ~35% hitting target. Crucially, these numbers swing hard with how strictly the pattern is defined: Bulkowski's high-and-tight flag (a much stricter variant — a near-doubling flagpole then a tight, brief pause) is one of his best performers (cited ~15% failure, ~39% average rise), whereas loosely-drawn flags perform little better than chance. So "flags are reliable continuation patterns" is true only for tightly-specified instances; the generic pattern's textbook reliability is overstated. Hand-drawn pattern recognition also injects survivorship and selection bias that inflate retrospective hit-rates relative to live trading.

Strengths & limitations

Works best when: there's a strong, clean prior trend; the consolidation is tight and short with visibly contracting volume; and the breakout comes on a clear volume expansion. The patterns add the most value as risk-defined setups — the pattern boundary gives an objective, nearby stop — rather than as directional oracles.

Fails when: the pattern is loosely drawn (the single most common misuse — fitting trendlines to noise), volume doesn't confirm, or the trader assumes the lore-direction (especially treating ascending = bullish-certain or descending = bearish-certain). Throwbacks/pullbacks are common (Bulkowski: ~62–65% for triangles), so an immediate re-test of the breakout level is normal and not by itself a failure. All figures are regime- and sample-dependent and have decayed over time.

System relevance

Sibling nodes: this is a continuation-pattern definition node — the swing-specific entry/stop/target/hold mechanics live in the Swing Trading branch (cross-link, not duplicated here), and it sits alongside the reversal-pattern node under Chart Patterns. For the Augustus trade-setup agent, the hard caveat to carry forward is: do not treat a triangle's name as a directional signal, and do not import generic flag/pennant "reliability" — gate any continuation-pattern input on (i) a confirmed breakout close, (ii) volume confirmation, and (iii) pattern tightness, and weight it by Cairn's measured track record rather than the textbook base rates above.

Sources

• Bulkowski, thepatternsite.com — Flags, Pennants, Ascending Triangles, Symmetrical Triangles, Descending Triangles, High & Tight Flag study (measured base rates; all figures attributed to Bulkowski's bulk-equity samples)
• StockCharts ChartSchool — Flag, Pennant (Continuation), Symmetrical Triangle, Ascending Triangle (structure, volume signature, measure rule)
• Disputed/decaying: Bulkowski flags performance decay over decades; hand-drawn pattern hit-rates carry selection bias and are not equivalent to live-trading expectancy.