Chart Patterns: Head and Shoulders, Double Tops, Triangles, Flags

There are no success rates in this guide. The percentages attached to chart patterns in most books and articles name no data set, no sample period, no definition of a valid pattern and no rule for what counted as a win, which leaves nothing in them that anybody can check. What a pattern does supply is three prices: where it breaks, how far the structure implies the move could run, and where the idea is finished. Those are measurements, and this guide works all three for each formation.

The one rule behind every measured move

Take the height of the pattern, measured from its extreme to its break level, and project that distance from the break level in the direction of the break. That is the whole method. It is the same for every formation below.

The reasoning behind it is loose and worth stating as loosely as it deserves: a structure that took a certain amount of disagreement to build tends to resolve over a comparable distance. Price frequently stops short of the projection. In strong trends it runs well past. Use the projection to decide whether the trade is worth taking at all.

Head and shoulders

Three peaks, the middle one highest. The neckline runs through the two intervening lows.

Left shoulder at 76, head at 82, right shoulder at 75, neckline at 70. Height is 82 - 70 = 12. A close below 70 projects to 70 - 12 = 58. Invalidation is a close back above 70 after the break, and the pattern is finished entirely on a close above the right shoulder high of 75.

Those two invalidation choices are a real decision with a price attached. Entering at 69.60 with a stop above 70.40 risks 0.80 per share against an 11.60 projection, while entering the same trade with the stop above 75.20 risks 5.60 per share for that projection, which turns a 14 to 1 structure into a 2 to 1 structure. The tighter stop is removed more often. Both are defensible and you have to pick one before you enter.

Underneath the name is a sequence of failures to make a higher high. That is a downtrend forming. That description is worth more than the anatomy. An inverse head and shoulders mirrors everything. Draw the neckline across the two intervening highs and add the projection above it.

Double tops and double bottoms

Two highs at approximately the same price with a pullback between them. The neckline is the low of that pullback.

Highs at 55.00 and 54.80 with an intervening low at 49.00. Height is 55.00 - 49.00 = 6.00, so a close below 49.00 projects to 49.00 - 6.00 = 43.00. Invalidation is a close above 55.00. The level has then been cleared, and the structure is a base.

Double bottoms reverse it. Lows at 31.00 and 31.20 with an intervening high at 35.00 give a height of 35.00 - 31.00 = 4.00 and a projection of 35.00 + 4.00 = 39.00 on a close above 35.00, invalidated on a close below 31.00.

The two highs do not need to match. Within roughly 2% of each other is a reasonable tolerance, and the second high coming in marginally above the first is common, because clearing the old high is what triggers the stops of everyone who shorted it.

Ascending and descending triangles

An ascending triangle is flat resistance with rising lows underneath it. Buyers are paying progressively higher prices into a fixed block of supply. A descending triangle is the reverse.

Worked, on an ascending triangle. Resistance is flat at 48.00. The first low is 42.00. The most recent higher low is 46.20.

Item	Arithmetic	Value
Widest height	48.00 - 42.00	6.00
Break level	Flat resistance	48.00
Projection	48.00 + 6.00	54.00
Entry	Above the break	48.30
Invalidation	Below the last higher low at 46.20	45.90
Risk per share	48.30 - 45.90	2.40
Reward on the projection	54.00 - 48.30	5.70
Ratio	5.70 / 2.40	2.4

Running that arithmetic before entry is the only reason to identify the pattern. A ratio of 2.4 is worth a look. The identical formation with resistance at 48.00 and a first low at 46.50 gives a height of 1.50, a projection of 49.50 and a reward of 49.50 - 48.30 = 1.20 per share. Against a stop below 46.20 that is under 1 to 1, and nothing on the chart tells you which of the two you are looking at, because the shape and the name are identical.

Symmetrical triangles contain no direction

Both boundaries converge, so the structure describes a contraction and nothing else. The break level is whichever boundary gives way, the projection is the widest height measured from the break, and the invalidation is a close back inside the triangle.

Traders assign a direction from the trend that preceded the contraction. That is a reasonable assumption and it remains an assumption, and it is worth saying out loud because the pattern is frequently presented as though the preceding trend were part of its definition.

Flags and pennants

A flag is a short consolidation after a sharp move, typically one to three weeks, drifting gently against the direction of that move. A pennant is the same pause in a small triangular shape.

The flagpole runs from 30.00 to 36.00, so the pole height is 36.00 - 30.00 = 6.00. The flag drifts back to a low of 34.50. Its upper boundary is 35.20. A close above 35.20 projects to 35.20 + 6.00 = 41.20. Invalidation is a close below the flag low of 34.50. A stop at 34.20 gives risk of 35.20 - 34.20 = 1.00 per share against a 6.00 projection.

The flag stops being a flag if the consolidation retraces most of the pole. Past roughly half of it, too much of the move has been given back, and what remains is an ordinary pullback, to be judged on levels.

The three numbers, collected

Pattern	Break level	Measured move	Invalidation
Head and shoulders	Neckline	Head to neckline height, projected down	Close above the right shoulder high
Inverse head and shoulders	Neckline	Head to neckline height, projected up	Close below the right shoulder low
Double top	The intervening low	Peak to low height, projected down	Close above the higher peak
Double bottom	The intervening high	High to trough height, projected up	Close below the lower trough
Ascending triangle	Flat resistance	Widest height, projected up	Close below the last higher low
Descending triangle	Flat support	Widest height, projected down	Close above the last lower high
Symmetrical triangle	Either boundary	Widest height, projected from the break	Close back inside the triangle
Bull flag	Upper flag boundary	Pole height, projected up from the break	Close below the flag low
Bear flag	Lower flag boundary	Pole height, projected down from the break	Close above the flag high

Everything in that table is a statement about geometry. None of it requires a claim about how often anything works, which is the point: the discipline survives without the statistics, and the statistics do not survive scrutiny.

Context, which is where the pattern gets its odds

Patterns sit inside a trend and inside a market. A bull flag in a stock above a rising 200-day average is a different proposition from the identical flag in a name that has lost a third of its value, and the moving averages guide covers how to establish that backdrop.

Volume supplies the participation half. Compare the breakout session against the 50-day average: 4.9 million / 1.8 million = 2.7 times normal means a lot of people acted at the new price, while 0.7 times normal means the level was cleared by almost nobody. The volume analysis guide covers the measurement and the exhaustion case.

Every pattern in this guide is built out of support and resistance zones. Draw those first, using the support and resistance guide. The bars at the turning points often show the evidence earliest, which the candlestick chart guide covers.

A failed pattern is information

When a double bottom breaks down and immediately reclaims the low, everyone who shorted the break is trapped and their covering supplies the fuel for the move back up. That is why failures often travel faster than the pattern would have.

The candlestick scanner locates the reversal bars that often mark the turns inside these structures, and the stock screener filters for stocks near 52-week highs with the volume and moving average conditions attached, which is how you find candidates without hunting for shapes. Read the technical analysis pillar for the workflow that puts trend before shape, and run the chart patterns and indicators quiz to check the measured move arithmetic.