Chart Patterns Tested on 790 Breakouts: Ascending Triangles Broke Down 70% of the Time, and 0 of 30 Runs Passed

Every trading book has the same page of chart patterns. The ascending triangle is bullish, the descending triangle bearish, the rising wedge bearish, the flag a pause before the trend goes on. Trade the breakout, put the stop on the other side, and the target is the height of the pattern.

Pattern indicators and pattern scanners promise to find these shapes for you. The question nobody answers with numbers is whether the breakouts they find go anywhere. I drew the patterns with rules instead of by eye and counted them on the same bars I used for the candlestick pattern test and the support and resistance test.

How do you find a chart pattern with rules?

• Swings: a swing high is a bar whose high is above the 3 bars on each side. It is only known 3 bars later, and the test only uses it from then on.
• Two lines: the upper line runs through the last two swing highs of a 30-bar window, the lower line through the last two swing lows. This is how most auto pattern indicators draw them.
• The pattern must be intact: no close since the first swing point may sit more than 0.25 ATR outside either line.
• Shapes: if the gap between the lines shrinks to 60% or less, it is a triangle (ascending, descending or symmetrical by the slopes) or a wedge (both lines pointing the same way). If the gap stays between 80% and 125%, it is a channel when it slopes, a rectangle when it is flat. A flag is a channel, rectangle or wedge that follows a move of at least 3 ATR, is no taller than 60% of that move and does not slope with it.
• Breakout: a candle closes beyond the line projected to that bar while the previous candle closed inside. Each shape counts once per 15 bars.
• Data: EURUSD, June 3 to August 26, 2026: 16,926 M5 bars and 5,642 M15 bars built from them. Gold (XAUUSD) over the same weeks: 13,736 M5 bars and 4,577 M15 bars.
• Cost: 1 pip per EURUSD trade from my spread and swap measurement, $0.30 an ounce per gold trade.

Which way did each pattern break?

EURUSD M5 and M15, gold M5 and M15. Rectangles (88) and flags (30) are left out here: a flag breakout is only counted in the direction of the flagpole.

The ascending triangle broke down in 54 of 77 cases. It did so on every chart: 31.8% up on EURUSD M5, 35.3% on gold M5, 20.0% on EURUSD M15, 18.2% on gold M15. The descending triangle broke up 81.5% of the time on EURUSD M5 and 75.0% on gold M15.

The pattern behind the whole table is simple geometry. Price broke through the line that was moving toward it. In an ascending triangle that is the rising lower line, in a rising wedge also the lower line, in a falling channel the upper one. The wedges agree with the textbook for the same reason the triangles disagree with it. A flat line waits for price, a sloping line walks into it.

Does a pattern breakout follow through?

Before any trade rules, the plainest question: after the close beyond the line, did price move 1 ATR further in the breakout direction before it moved 1 ATR back, within 48 bars? A random bar with a random direction does it about half the time.

Breakouts in brackets. If both levels were touched inside one bar, the bad side counted first.

All four charts came in under the random bar. The wedge was the only shape that reached a coin flip on three charts. Within 3 bars, 41.9% to 54.8% of all breakouts had closed back inside the pattern, which is the false breakout every pattern trader knows from experience.

How often did the textbook target get hit?

The textbook trade puts the stop beyond the opposite line and aims for the height of the pattern. I measured how often price reached that target before that stop, within 96 bars.

The triangle target is the height at the wide end, a median 4 to 5 ATR, while its stop sits close because the lines have already narrowed. That target came first in about one triangle of four. Channels and rectangles have a smaller height, a median 2.3 to 3.0 ATR, and reached it about half the time. The small M15 rectangle cells have 4 to 7 trades and say little.

Did any pattern make money as a trade?

Per trade, trades in brackets. One trade at a time, so fewer trades than breakouts.

Then I split the result by shape. Every shape with at least 30 trades on a chart, times three exits, gave 30 runs. Each run was checked three ways: profit after cost, profit in both halves of the data split at July 15, and better than the best 5% of 300 runs of random entries with the same exit and the same number of trades.

The best of them was the channel breakout on gold M5 with a 12-bar exit: +$1.22 an ounce a trade over 73 trades, +$1.25 in the first half and +$1.21 in the second. Random entries reached +$2.01 in their best 5% with that many trades. The triangle on gold M5 made +$1.48 a trade, all of it in the first half. On EURUSD the only shapes in profit were the wedge on M5, by 0.23 pips a trade, and the rectangle on M15, on just 7 trades.

Gold fell about 10% to late June, ranged through July and rallied in August. A rule that catches only some swings still had big moves to catch, which is why the few positive rows are all gold.

So are chart patterns useless?

• Not as a picture of the market. A pattern tells you where the recent swings sit and how tight the range has become.
• Not as a breakout signal on M5 or M15 by themselves. On all four charts the breakout followed through less often than a random bar.
• Do not trust the textbook direction of a triangle. Watch the sloping line. On these bars it broke 70% of the time in ascending triangles.
• Do not use the full pattern height as the target of a triangle. It was hit first in about one case of four.

Whether a pattern does any better than a plain range breakout of the same 30 bars is in chart patterns vs range breakout. Which lookback, touch tolerance and number of touches worked, and how to draw the lines in the platform, is in the chart pattern indicator for MT4 guide. Breakouts measured by volatility instead of shape are in the ATR breakout test.

Limits of this test

June 16 to August 26, 2026 after the warm-up bars, about ten weeks, two markets and two timeframes. Patterns on H4 and daily charts, where many traders draw them, would need years of data to give enough breakouts. The 30-bar window, the 3-bar swing and the 60% narrowing are one set of choices; other windows and tolerances are in the settings guide. A trader drawing by eye picks the cleanest shapes and skips ugly ones, and no rule can copy that judgement. What a rule can do is show what the average shape does, and here the average shape did worse than a coin.

Frequently asked questions

Do chart patterns work in forex?

Not as a breakout rule on M5 and M15 in this test. Across 790 breakouts on EURUSD and gold, the breakout reached +1 ATR before -1 ATR 40.7% to 47.3% of the time, against 49.2% to 50.3% for a random bar, and 0 of 30 trading runs passed every check.

Is an ascending triangle bullish?

Not on these bars. 77 ascending triangles broke up only 29.9% of the time, and below 36% on every chart. The rising lower line met price first.

Which chart pattern is the most reliable?

The wedge matched the textbook direction best: rising wedges broke down 82.9% of the time and falling wedges broke up 77.3%. It was also the only shape near a coin flip on follow-through, which still is not an edge.

How often do chart pattern breakouts fail?

Within 3 bars, 41.9% to 54.8% of breakouts had closed back inside the pattern, depending on the chart.

Does the measured move target work?

For channels and rectangles about half the time. For triangles the full height was reached before the stop only 20.9% to 26.1% of the time.