Chart patterns: measurement and invalidation

What a chart pattern is, and what it is not

This lesson is for someone who can already name market structure and tell a trend from a range, and who now wants to translate chart shape into numbers: how many candles a pattern needs, exactly where the neckline runs, and where the target move comes from. The point is not to collect names, but to build definitions that two people can apply to the same chart and get the same result.

A pattern organises price behaviour around three points: where price compressed, where the exit from that compression lies, and where the thesis stops being true. Without the third point there is no setup — only a drawing. On liquid blue chips on the Warsaw Stock Exchange and on index CFDs, the shapes tend to be cleaner than on a thin small cap, where spread and gaps can destroy a textbook outline. Volume adds context on session-based equities; on many CFDs, treat it with caution.

False breakouts from a pattern are a normal part of the game, not an exception. That is why the order is always the same: definition and invalidation first, then position size, and only then the name of the pattern in your journal.

How to measure a pattern so that two people get the same result

Measuring a pattern consists of six elements. You record each of them as a number or as a specific level — never as “somewhere around here”. All the values in the block below are illustrative numbers: they are a convention you may change, as long as you keep it unchanged throughout the test.

H             = |neckline − pattern_extreme|              (pattern height)
bar_min       = number of candles from the first extreme to the breakout candle
breakout      = a candle close beyond the neckline (or the wick alone — choose one)
target        = neckline ± k × H                          (illustrative number: k = 1.0)
invalidation  = the level on the opposite side of the pattern
pattern_RR    = (k × H) / |entry − invalidation|

The steps, in this order:

1. Write the extremes down as numbers. The three peaks of a head and shoulders and the two tops of a double top each have a specific value. “Around 1820” is not a measurement.
2. Define the neckline as a level. It is a straight line drawn through the two troughs (or two tops) that separate the extremes of the pattern. You record it as a value, not as “sort of here”.
3. Calculate the pattern height H. It is the distance from the neckline to the most extended extreme.
4. Calculate bar_min — the number of candles from the first extreme to the breakout candle. If the pattern is shorter than the minimum adopted for its type (illustrative number: see the pattern table below), there is nothing from which to calculate a target.
5. Write down the breakout condition: a close beyond the neckline, or a wick being enough. One decision for the whole test.
6. Define the invalidation on the opposite side of the pattern, before you enter.
7. Calculate the target and pattern_RR, and record the result in R-multiples together with the measurement.

Head and shoulders: a measurable definition

A reversal pattern in which the middle peak is the highest and the two outer ones are close to each other. To be comparable across instruments it needs two tolerances expressed relative to the pattern height:

H            = |head − neckline|
neckline     = a straight line through the two troughs between the shoulders
tolerance_1  = |trough_1 − trough_2| / H                (illustrative number: ≤ 0.15)
tolerance_2  = |left_shoulder − right_shoulder| / H     (illustrative number: ≤ 0.30)

If tolerance_1 is exceeded, the two troughs do not lie on one sensible line and the pattern stops being a head and shoulders. This is not a matter of taste — it is a verifiable condition. Only once it holds do you calculate the target as a multiple of H and compare pattern_RR with the threshold in your plan. A shape whose tolerances fail is described in your journal as a “rejected pattern”, not as “almost a head and shoulders”.

Double top and double bottom: a measurable definition

The same logic with fewer elements. Two tops at a similar level, one trough between them, and the pattern height is the distance from the top to that trough.

H            = |top_1 − trough_between_tops|
tolerance    = |top_1 − top_2| / H              (illustrative number: ≤ 0.10)
min_bars     = number of candles between the tops   (illustrative number: ≥ 5, the length of the consolidation)
target       = trough − k × H                   (for the top variant)
invalidation = above the higher of the two tops

Two tops two candles apart are usually one impulse with a correction, not a reversal pattern — which is why min_bars is part of the definition rather than a detail. min_bars describes only the distance between the extremes; you measure the total length of the pattern with a separate parameter, bar_min (illustrative number: ≥ 10), and both numbers must hold. For the bottom variant you flip the signs: the target is measured upwards from the top between the troughs, and the invalidation is set below the lower of the troughs.

Always calculate the target from the neckline — for a double top that is the trough between the tops, and for a double bottom the top between the troughs. Reading it from the second extreme gives a result that differs by a full H, so the two methods are not interchangeable and cannot sit side by side in one definition.

Flag and triangle: compression measured in candles

Continuation patterns describe a pause in a move, not its reversal. A flag and a triangle differ in geometry, but the measurement stays the same: the length of the consolidation, the height of the impulse and the point of the breakout.

impulse       = the preceding move ≥ 2 × ATR(TF, n)        (illustrative number, n = 14)
consolidation = 5–15 candles whose combined range does not exceed 1/3 of the impulse height
triangle      = range of the last 5 candles / range of the first 5 candles  (illustrative number: ≤ 0.6)
breakout      = a close beyond the boundary of the consolidation
target        = the breakout point ± the height of the consolidation

ATR(TF, n) is the average True Range over n periods; take n = 14 (an illustrative number) and keep that value unchanged for the whole test. You will find the definition of ATR in the lesson Multi-timeframe — top-down analysis; here it serves only to tell a real impulse from noise. A flag made of three candles is not compression — it is a pause that ordinary noise fills in.

The bar_min column counts the whole pattern, while the consolidation-length column counts only its compressed part. With a flag and a triangle the two numbers may differ by several candles, because bar_min also covers the impulse that precedes the consolidation.

The multiplier k deliberately does not appear in that table — it belongs to your plan. A value of 1.0 is often a starting point, 0.5 gives a target that is easier to reach, and 1.5 requires a move that appears rarely on many instruments. Check on your own sample at which k the target is at all reachable before invalidation — and only then fix it for good.

Worksheet: fill it in with your own data

Copy the table into a spreadsheet and complete it before every entry. The columns are empty on purpose — what has value is what you measure on your own chart. You fill in the last column after exiting the position.

After a dozen rows, compare the patterns that met the tolerances with those that exceeded them. If the results of the two groups are similar, the tolerances are not doing the job of a filter — and that is the correct conclusion from the exercise, not a reason to loosen them. The lesson stop-loss and R-multiple explains how to record the position size that follows from the distance to invalidation.

When a pattern says “do not trade”

Four conditions, each stated as a rule. Meeting any one of them means skipping the setup — not “entering with a smaller position”.

1. bar_min below the minimum for that pattern (illustrative numbers from the pattern table: 20 for head and shoulders, 10 for a double top, a flag and a triangle). For a flag and a triangle, bar_min counts the whole pattern — the impulse together with the consolidation — so the consolidation itself may be 5 candles long. A shorter pattern is usually a single candle of noise, from which no target can be calculated repeatably.
2. A breached neckline or symmetry tolerance. If the troughs do not lie on one sensible line, the pattern does not meet the definition. Every time you shift the neckline “so that it fits”, you create a new pattern — and you must measure that one from scratch as well.
3. A breakout by wick when the plan requires a close. The breakout condition is part of the definition, not decoration. Price can still return to the consolidation while the candle is forming.
4. pattern_RR below your plan’s threshold after costs (illustrative number: threshold 2.0). On a CFD, a tight target is often eaten by spread and slippage; you check the cost before the decision, not after the loss.

The order of operations is the opposite of intuition: first you reject the patterns that fail the definition, then you plan the ones that meet it. A rejected pattern is a normal work result, not wasted analysis. You apply the same order to indicators: first you write down the period and the price source, and only then do you read the signal — the lesson MA, RSI and MACD indicators shows how to build such a protocol.

Common mistakes when working with patterns

• Collecting names instead of measuring one or two patterns across many sessions.
• Mixing frames of reference — height measured from a wick and the breakout from a close. Rows like that in a journal are not comparable.
• Treating the target as a promise — a multiple of the height is a hypothesis to test, not a forecast.
• Measuring only successful examples — a sample made up entirely of successes says nothing about your definition.
• Ignoring zone context — a pattern with no reference to support and resistance easily turns into a fit to history.
• Treating Fibonacci retracements as an oracle — 38.2% or 61.8% is a zone to check in the context of the trend, not a level that “must” hold. A level without confirmation and without invalidation stays a drawing, not a setup.
• Changing the definition during the test — loosening the tolerances after a losing streak invalidates the whole sample collected so far.

Measurable exercise

Choose one instrument and one timeframe. Copy the definition of one pattern — a double top, for instance — into your plan along with the tolerance, the minimum candle count and the target multiplier. Then find 20 past formations that look like that pattern, and for each one calculate H, bar_min, the tolerance and pattern_RR.

Write down how many formations met the tolerance and how many exceeded it. The goal is not to calculate how effective the pattern is, but to check whether your definition gives the same result on a second pass through the same chart. If after a week you measure H differently than you did on day one, the problem is in the definition, not in the market. The lessons defining a setup and backtesting and forward testing describe how to run such a sample.

Common questions

How many patterns do I need to master at the start?

One or two — a double top and a continuation flag, for example. Measure them on your own sample and record the results in R; further names will add nothing until those two can be measured repeatably.

Does volume confirm a pattern?

On markets with reliable volume, such as Warsaw Stock Exchange equities, it can be useful context at the breakout. On many CFDs, volume is only an approximation of turnover from the underlying market, so do not build an entire definition on it. Check it on your own instrument and write the conclusion into your plan.

What should I do about a false breakout from a triangle?

It is a normal event, not a breakdown of the pattern. Either you trade the return into the consolidation after the breakout, or you skip the entry when there is no close beyond the line — as long as the choice was written down before the session, not after a loss.

Why measure a pattern when the shape is visible on the chart?

Because 'visible' means something different to two people looking at the same chart. Numbers — the pattern height, the candle count in the consolidation, the position of the neckline — let you record a measurement in your journal and compare it across sessions and instruments. Only from recorded values can you work out whether your tolerances genuinely filter out worse formations or merely sound precise. A shape without measurement stays an impression that you can neither improve nor defend to yourself after a losing streak.