Flags and pennants — continuation patterns after strong moves

Flags and pennants describe a consolidation after a strong price move. The flag uses a roughly parallel channel; the pennant uses converging boundaries. Both are commonly classed as continuation patterns, but a shape does not specify who traded or guarantee the next move. This guide turns the description into dated geometry, an explicit breakout rule and a calculation from actual entry and exit prices.

What flags and pennants are — twin continuation patterns

The preceding impulse is the flagpole, which can point upward or downward. A flag’s boundaries are approximately parallel; a pennant’s range narrows between converging lines. Record the preceding impulse as well as the consolidation: a small triangle alone does not define a pennant. Any numerical duration or size threshold belongs to the particular detection rule being tested.

CME presents flags and pennants as possible clues to continuation and explicitly says they do not always signal it. A flagpole-height projection is a reference construction, not evidence of success on a currency pair. Geometry differs, so the rules for drawing boundaries and measuring a break must also be recorded.

The usual continuation interpretation treats the consolidation as a pause after an impulse. That is a hypothesis about subsequent price movement, not evidence that institutions are completing orders. The pattern can fail, and its flagpole projection is not a guaranteed follow-through distance.

The mechanics of a flag — a parallel corrective channel

A flag has the geometry of a parallel channel in which the two boundary lines run at the same angle, usually tilted against the direction of the flagpole. After a strong rally the flag channel drifts gently downward; after a sharp decline it drifts gently upward. Visually the pattern resembles a small flag hanging on a pole — which is exactly where the name comes from.

Define the start and end of the pole, the points used for each boundary and the permitted line-fitting tolerance. Choose whether a break means an intrabar price breach or a completed close beyond the line. A two-ATR pole, 50% retracement or fixed pip buffer can be an explicitly chosen test parameter; none is established here as a universal credibility threshold.

ATR averages true range, which includes differences from the preceding close as well as High–Low; it is not simply average daily High–Low. Record its lookback, calculation and data source if using it to scale a detector. Consolidation length alone does not reveal institutions refilling positions or establish a sharper future move.

The mechanics of a pennant — a small converging triangle

A pennant differs from a flag in the geometry of its consolidation: instead of a parallel channel it draws a miniature symmetrical triangle on the chart. Two converging trendlines close in at similar angles — the highs are progressively lower and the lows progressively higher — and the price range systematically tightens. Visually the pattern looks like a small triangular flag wrapped around the pole, which is exactly where the word "pennant" comes from.

A pennant’s two projected boundaries also have a finite geometric domain. In an original hypothetical example, using arbitrary price units, U(t) = 120 − 0.5t and L(t) = 110 + 0.25t are fitted from points at t = 0 and t = 8. Width is 10 − 0.75t; the apex is t = 13⅓. Widths at t = 0, 8 and 12 are 10, 4 and 1; at t = 14, U = 113 and L = 113.5 give −0.5. The formerly upper line is now below the lower one.

Extending those lines does not extend the original region indefinitely. A rule allowing observations after their intersection needs its own definition. The apex follows from the selected points; it is not a deadline by which price must accelerate or reach a projection.

Calling both patterns continuation formations does not make their portfolio outcomes equivalent. A detector can use different boundaries, timing and invalidation rules for each. Compare all qualifying cases under a fixed definition, including false breaks and cases with no order, instead of transferring a stock-pattern claim to OTC forex.

The flagpole as a yardstick — the foundation of the entire pattern

The flagpole supplies a height for one projection convention. In a hypothetical EUR/USD model, a Bid move from 1.1000 to 1.1200 has height 0.0200 USD/EUR, or 200 pips at 0.0001 per pip. Specify which start and end points were available when you measured it; the drawing does not establish a statistical edge.

For this fictional flag, upper points (t = 4, 1.1200) and (t = 8, 1.1180) define a decline of 0.0005, or 5 pips, per bar. Lower points (t = 6, 1.1130) and (t = 10, 1.1110) have the same slope. Assume they were manually chosen and frozen by the close of bar 10; this is selected geometry, not a complete OHLC series or historical trade.

The channel is consistently 0.0060, or 60 pips, wide, so it is not a pennant. At t = 11 the upper boundary is 1.1165 and Bid closes at 1.1160: no signal under the chosen Close > U rule. At t = 12 the boundary is 1.1160 and Bid closes at 1.1170: the condition passes then. A swing selected using future bars would need a later availability time.

Volume: identify the dataset and test the filter

Some technical-analysis descriptions suggest increasing volume on the impulse and breakout and declining volume during consolidation. Treat that as a proposed filter, with its source and units stated, rather than a signature identifying the participants. Neither a volume bar nor a candlestick reveals an institution’s remaining order.

Percentages such as 150%, 200% or 300% of a volume average are possible chosen thresholds, not proven universal requirements here. Define the denominator’s lookback, comparable sessions and missing-data handling before reviewing outcomes. Report failed breakouts as well as completed projections; removing failures after seeing the future changes the evaluated rule.

MT5 distinguishes received-tick count from real deal volume, which may be unavailable for OTC instruments. A broker’s stream does not aggregate every spot-FX trade. Currency-futures volume describes its own listed contract and venue; it is not the same dataset. The OBV (On-Balance Volume) indicator transforms its input volume and cannot convert tick counts into global notional turnover or identify a large-player breakout.

Entry rules, stop loss and position management

Compare entry variants as separate protocols. Waiting for a close changes when information is available; a retest may never occur; an intrabar rule uses a different condition. No variant is proven here to offer the safest or best execution. Record the instrument’s price side, order type and any buffer before testing.

1. Entry after a completed close. Apply the stated closing condition only when the candle ends. The next executable price can differ from its close. A completed break can still fail; this timing choice does not guarantee a lower loss or a profitable trade-off.
2. Entry conditional on a retest. Define the allowed return, confirmation and expiry. Price may continue without filling the order, or return and invalidate the structure. Assess the resulting fills and missed trades instead of assuming a better price and tighter risk.
3. Intrabar breach. Specify the live price condition and order trigger. A brief breach can pass this rule while failing a closing rule, as the diagram illustrates. A ten-pip buffer would be a model choice, not guaranteed protection or the best possible fill.

For a bullish breakout, a possible invalidation level is below the lower boundary or last consolidation swing low; the bearish case is reversed. A stop beyond the opposite boundary will not trigger on every return inside the pattern. Choose any buffer from the instrument's volatility and execution conditions; a fixed five- or ten-pip buffer does not prove deliberate stop hunting or guarantee protection.

For a hypothetical USD 10,000 account and USD 100 planned price-loss budget, a 40-pip distance permits USD 2.50 per pip before other costs. On an assumed EUR/USD contract with USD 10 per pip per standard lot, that is 0.25 lot. A EUR account needs the actual USD-to-EUR conversion first. Costs, rounding and slippage can change the realized loss.

A split exit is another testable variant: specify the partial volume, exit condition and rule for the remainder. A trailing order or moving-average rule does not guarantee protection at its displayed level. Compare its complete realised cash flows with an unsplit variant on the same cases, including costs.

Price target projection — the flagpole height as reference

The classical price target after a breakout from a flag or a pennant follows directly from the geometry of the pattern. The price projection target equals the flagpole height projected from the breakout point in the direction of the move. Flagpole height is measured as the vertical distance between the start of the sharp move (the point where volume and volatility began to expand) and the peak of that move — or the trough for a bearish flag.

In the fictional 200-pip example, project the pole from the frozen upper boundary at t = 12: 1.1160 + 0.0200 = 1.1360. This is a reference level, not an entry or promised destination. Suppose a later buy actually fills at Ask 1.1173 for EUR 20,000, with a USD account and total commission USD 4. One pip is USD 2. An exact exit at Bid 1.1360 gives USD 374 gross and USD 370 after commission.

An alternative protective exit at Bid 1.1105 gives −USD 136 before commission and −USD 140 after it. If the fill is instead 1.1102, the result is −USD 146 after commission. Ask/Bid prices already contain the spread; the model excludes funding and other fees and needs no currency conversion. The projected 200 pips are not 200 pips of gain from the actual entry.

Alternative projections need named anchors. A 100 percent Fibonacci extension can start from a different point than the breakout boundary, so agreement should be calculated rather than assumed. An ATR multiple is not a ceiling on future price movement. In the same fictional trade, an entry at Ask 1.1183 instead of 1.1173 makes an exact Bid 1.1360 exit USD 350 net and Bid 1.1105 exit −USD 160 net: ten pips worse at entry changes each result by USD 20.

Five mistakes that destroy flag and pennant trading

Keep the chart description, signal time and execution calculation separate. The following checks identify definitions that would otherwise change during a test; none establishes a trading advantage merely by being followed.

• Omitting the impulse definition. Record the pole and boundary-selection rules. A chosen ATR threshold can distinguish cases in your test, but it is not established here as the universal definition of a genuine flag.
• Confusing tick activity with turnover. Record the feed, unit and averaging window. A higher tick count does not identify institutions or validate a breakout by itself.
• Changing the higher-timeframe filter. If a trend filter belongs to the rule, define it before looking at results and use information already available. Opposing directions on different horizons do not reveal which participants traded.
• Mixing breach and close rules. A wick beyond a boundary is not a completed close beyond it. Keep their signal times separate; a return inside does not prove deliberate stop hunting.
• Assuming a universally best timeframe. A different bar interval changes aggregation, costs and available observations. Define the interval and test it; no evidence here makes M5 or M15 worthless or H4 intrinsically profitable.

What to do tomorrow

1. Specify the detector. Record price source, timeframe, pole points, boundary fit, tolerances and signal availability. If using ATR, distinguish true range from High–Low and identify the selected multiplier as a test parameter.
2. Preserve every qualifying case. Record breaks, failed breaks and no-fill scenarios while keeping future bars hidden during identification. Then evaluate the stated entry/exit rules using executable prices and costs; selected clean charts do not prove performance.
3. Write down your entry rule and stick to it for the next month. Pick one of the three variants described above — on the candle close, on the pullback to the line, or during the breakout itself — and do not switch mid-series. Comparing results only makes sense when you did the same thing throughout the test.

Related reading: triangles — continuation patterns in technical analysis to understand how full triangles differ from miniature pennants; pennant pattern — bull/bear flag continuation for a concise take on pennant strategy; wedge pattern — rising/falling wedge reversal strategy to see how wedges differ from flags and pennants.