Chart patterns
Continuation patterns: 39 chart and candlestick patterns, tested
A continuation pattern is a pause inside a trend: price moves sideways or against the trend for a while, then, textbooks say, resumes in the same direction. Flags, pennants, triangles, rectangles and the cup and handle are the classic chart examples; rising and falling three methods are the best-known candlestick ones. We tested all 39 on the full price history of 2,368 assets across six markets, against random bars of the same assets.
Statistics updated November 5, 2026, from the latest run of our test bench. Historical statistics before fees and slippage. Not investment advice: past results do not predict future ones.
- Continuation patterns tested
- 39
- Occurrences detected
- 369,227
- Beat chance after all checks
- 0
- Moved opposite to the textbook
- 3
The short answer
None of the 39 continuation patterns beat chance after all our checks (multiple-testing correction, stability over time, costs), neither on all markets pooled nor on any single market and timeframe.
Strongest result on its largest sample: Symmetrical triangle, promising on the 2:1 trade test, price moved the expected way within 10 bars 50.5% of the time against 50.6% for random bars (12,937 occurrences, all markets, 1-hour).
8 patterns on this page (symmetrical triangle, rectangle, descending triangle, ascending triangle, bear pennant, bear flag, bull flag and bull pennant) can break either way, so we test the breakout in each direction. The continuation idea itself (that the break follows the prior trend) is a claim about direction: where our verdict says "No clear edge", the pattern did not pick the direction better than chance.
Chart continuation patterns 21
Patterns drawn over dozens of bars: a pause between two legs of the same trend. Entry on the breakout close, stop on the other side of the pattern, target at twice the risk, in our 2:1 trade test.
- Symmetrical trianglePromising 2:1 tradeExpected move after 10 bars: 50.5% vs 50.6% at randomdownward breakouts; upward: same verdict12,937 occurrences, all markets, 1-hour
- RectanglePromising 10-bar moveExpected move after 10 bars: 50.6% vs 50.6% at randomdownward breakouts; upward: No clear edge4,399 occurrences, all markets, 1-hour
- Bearish J-hookOpposite of the textbook 10-bar moveExpected move after 10 bars: 47.1% vs 50.5% at random106,578 occurrences, all markets, 1-hour
- Measured move downOpposite of the textbook 10-bar moveExpected move after 10 bars: 45.5% vs 50.6% at random43,066 occurrences, all markets, 1-hour
- Measured move upOpposite of the textbook 10-bar moveExpected move after 10 bars: 45.3% vs 48.7% at random38,962 occurrences, all markets, 1-hour
- Bullish J-hookNo clear edge 10-bar moveExpected move after 10 bars: 46.8% vs 48.8% at random80,065 occurrences, all markets, 1-hour
- Descending triangleNo clear edge 10-bar moveExpected move after 10 bars: 47.3% vs 48.4% at randomupward breakouts; downward: Promising10,822 occurrences, all markets, 1-hour
- Flat top breakoutNo clear edge 10-bar moveExpected move after 10 bars: 46.5% vs 48.7% at random10,620 occurrences, all markets, 1-hour
- Volatility contraction patternNo clear edge 10-bar moveExpected move after 10 bars: 54.1% vs 54.6% at random9,220 occurrences, all markets, daily
- Ascending triangleNo clear edge 10-bar moveExpected move after 10 bars: 50.8% vs 50.5% at randomdownward breakouts; upward: same verdict8,524 occurrences, all markets, 1-hour
- Runaway gap upNo clear edge 10-bar moveExpected move after 10 bars: 49.6% vs 54.7% at random3,420 occurrences, all markets, daily
- Runaway gap downWeak, below chance 10-bar moveExpected move after 10 bars: 48.7% vs 45.8% at random2,010 occurrences, all markets, daily
- Bear pennantNo clear edge 10-bar moveExpected move after 10 bars: 50.1% vs 50.7% at randomdownward breakouts; upward: Weak, below chance1,891 occurrences (undercounted), all markets, 1-hour
- Bear flagNo clear edge 10-bar moveExpected move after 10 bars: 49.4% vs 50.7% at randomdownward breakouts; upward: same verdict1,736 occurrences (undercounted), all markets, 1-hour
- Bull flagNo clear edge 10-bar moveExpected move after 10 bars: 49.8% vs 48.8% at randomupward breakouts; downward: same verdict1,245 occurrences (undercounted), all markets, 1-hour
- Inverted cup with handleNo clear edge 10-bar moveExpected move after 10 bars: 47.8% vs 50.3% at random1,218 occurrences, all markets, 1-hour
- Bull pennantNo clear edge 10-bar moveExpected move after 10 bars: 44.7% vs 48.8% at randomupward breakouts; downward: same verdict1,192 occurrences (undercounted), all markets, 1-hour
- Cup with handleNo clear edge 10-bar moveExpected move after 10 bars: 46.6% vs 48.6% at random1,186 occurrences, all markets, 1-hour
- Ascending scallopNo clear edge 10-bar moveExpected move after 10 bars: 43.3% vs 48.6% at random892 occurrences, all markets, 1-hour
- Descending scallopNo clear edge 10-bar moveExpected move after 10 bars: 43.6% vs 50.3% at random675 occurrences, all markets, 1-hour
- High and tight flagNo clear edge 10-bar moveExpected move after 10 bars: 49.0% vs 51.8% at random304 occurrences, all markets, daily
Continuation or reversal: the wedges 2
Textbooks list wedges as both: a falling wedge inside an uptrend is a continuation, at the end of a downtrend a reversal. Our rule does not depend on the prior trend, so the result below covers both readings.
- Falling wedgeNo clear edge 10-bar moveExpected move after 10 bars: 50.4% vs 48.4% at randomupward breakouts; downward: same verdict10,673 occurrences, all markets, 1-hour
- Rising wedgeNo clear edge 10-bar moveExpected move after 10 bars: 50.5% vs 50.3% at randomdownward breakouts; upward: same verdict7,103 occurrences, all markets, 1-hour
Candlestick continuation patterns 16
Two to five candles inside a trend, read as a short rest before the next leg. Our rules require the trend to be there first, as textbooks do.
- ThrustingNo clear edge 10-bar moveExpected move after 10 bars: 45.2% vs 45.6% at random6,762 occurrences, all markets, daily
- On neckNo clear edge 10-bar moveExpected move after 10 bars: 49.1% vs 45.8% at random742 occurrences, all markets, daily
- In neckNo clear edge 10-bar moveExpected move after 10 bars: 47.1% vs 45.6% at random681 occurrences, all markets, daily
- Upside tasuki gapNo clear edge 10-bar moveExpected move after 10 bars: 52.1% vs 55.0% at random560 occurrences, all markets, daily
- Downside tasuki gapNo clear edge 10-bar moveExpected move after 10 bars: 41.4% vs 45.3% at random364 occurrences, all markets, daily
- Up-gap side-by-side white linesNo clear edge 10-bar moveExpected move after 10 bars: 56.5% vs 55.0% at random249 occurrences, all markets, daily
- Bullish separating linesNo clear edge 10-bar moveExpected move after 10 bars: 56.5% vs 54.7% at random240 occurrences, all markets, daily
- Bullish three-line strikeNo clear edge 10-bar moveExpected move after 10 bars: 52.0% vs 54.0% at random221 occurrences, all markets, daily
- Bearish three-line strikeNo clear edge 10-bar moveExpected move after 10 bars: 45.6% vs 46.4% at random158 occurrences, all markets, daily
- Bearish separating linesNo clear edge 10-bar moveExpected move after 10 bars: 42.6% vs 45.2% at random148 occurrences, all markets, daily
- Rising three methodsNo clear edge 10-bar moveExpected move after 10 bars: 50.5% vs 53.2% at random108 occurrences, all markets, daily
- Falling three methodsNo clear edge 10-bar moveExpected move after 10 bars: 49.4% vs 47.0% at random82 occurrences, all markets, daily
- Downside gap three methodsNo clear edge 10-bar moveExpected move after 10 bars: 53.0% vs 45.0% at random67 occurrences, all markets, daily
- Upside gap three methodsNo clear edge 10-bar moveExpected move after 10 bars: 53.6% vs 54.4% at random56 occurrences, all markets, daily
- Mat holdNo clear edge 10-bar moveExpected move after 10 bars: 55.6% vs 54.6% at random36 occurrences, all markets, daily
- Down-gap side-by-side white linesNot enough data 10-bar moveExpected move after 10 bars: 40.0% vs 45.0% at random15 occurrences, all markets, daily
Each card: the shape, our verdict on the pattern's largest sample, and how often price moved the expected way within 10 bars against random bars of the same assets. A green border marks a pattern that beats chance after all our checks, on the scope named on the card. Promising: better than chance on one of our two tests, but it did not pass all our checks. It can still be luck. * Undercounted: the count is what our strict rule detected, not every occurrence that formed. †Unverified: the detection was checked on too few cases to be confirmed, so read the whole line with caution. ‡ XABCD groups four harmonic patterns (Gartley, bat, butterfly, crab) that are also counted on their own, so it is not a separate result. Before fees; past behavior, not a promise.
How to trade a continuation pattern
- Start with the trend. A continuation pattern is only defined inside a trend. If you cannot say which way the market was going before the pause, there is nothing to continue.
- Trade the breakout, not the shape. The pattern is a promise until price closes outside it. Entering inside the pattern means betting on a direction the pattern has not given yet.
- Put the stop on the other side. If price breaks out and comes back through the opposite side of the flag, triangle or rectangle, the pattern has failed. That is where our 2:1 test puts its stop.
- Know your target before you enter. The classic target is the height of the pattern, or of the pole for a flag, added to the breakout. Our test uses a fixed 2:1 target instead, so every pattern is judged on the same trade. Example: the bull flag, no clear edge on the 10-bar move test, price moved the expected way within 10 bars 49.8% of the time against 48.8% for random bars (1,245 occurrences, all markets, 1-hour).
How we tested
Each pattern is a strict written rule, run on the full price history of 2,368 assets across six markets, using only what was known at the time. What followed each occurrence is compared with what follows random bars of the same assets, on two tests: the move over the next 10 bars against the rest of the market, and a 2:1 trade against the same trade on random dates. A result only counts once it survives a correction for the thousands of tests we run, holds in both halves of the history and outlasts estimated costs. The whole pipeline first runs on random-walk prices, where no pattern can work. Method and data: the full study and open data.