Double top and double bottom: geometry, tolerance and failed breakouts

Consider a hypothetical daily EUR/USD structure with two nearby peaks and an intermediate trough. A short-entry variant requires a candle close below that trough; its stop and measured target are separate rules. This is a model for explaining double tops and bottoms, not a verified trade or chart from 3 March 2025.

What the M and W patterns are and where they come from

A double top, or M, is usually studied after an upward move: two similar highs are separated by a trough whose low defines the neckline. A double bottom, or W, uses two similar lows and the intervening high after a downward move. “Reversal pattern” names an interpretation, not a known future event. A historical book or an equity-sample statistic is not a global FX performance guarantee.

Separate the candidate shape from completion under a chosen rule and from a later outcome. Similar highs or lows alone do not complete the structure. TradingView’s documented detectors use a close beyond the intermediate level for the formed status; that status does not itself specify an order, size or fill. A similar shape in a lateral range may simply test its boundaries, so define the preceding context.

The chart cannot identify whose conviction changed or whether anyone deliberately triggered other traders’ orders. A neckline break does not prove universal surrender by buyers or sellers. If a strategy adds contextual or activity conditions, define them before the outcome and evaluate them separately instead of explaining each finished chart with a causal story.

The mechanics of the double top — the anatomy of the letter M

Define how similar the two extrema must be. A difference in pips, as a fraction of pattern height, or relative to another volatility measure uses a different denominator. TradingView’s permissible-deviation descriptions refer to heights or depths relative to the intermediate point, not simply a percentage of the nominal exchange rate. An arbitrary compulsory number of daily weeks is not part of our exercise.

For M, difference 0.0008 is 10% of H, but only about 0.0714% of the first high 1.1200. A nominal-price allowance of 3% at 1.1200 would be 0.0336, or 336 pips, more than our 80-pip pattern height. Reporting a “percentage difference” without its denominator can therefore select a very different set of shapes.

Record spacing and the local-extremum detector as explicit parameters rather than conclude that no rule can be written. Pivot recognition may need later bars; retrospective markers must not be treated as information known at the pivot time. The documented TradingView scanner uses its own 5/5 pivots, which our already-recognized-point model does not reproduce.

The mechanics of the double bottom — the anatomy of the letter W

The W counterpart reverses the geometry: its intermediate high gives the neckline, and a selected close above it can complete the pattern under a stated rule. That does not establish a subsequent rise. In the teaching W model, changing only the second low from 1.0812 to 1.0813 keeps H = 0.0080 but changes difference/H to 0.0013/0.0080 = 16.25%, so it fails our 15% limit regardless of what happens later.

Reciprocal quotation can change this same tolerance. For hypothetical highs 100 and 99.23 with neckline 95, H = 5 and difference/H = 0.77/5 = 15.4%, so M fails. Under 1/price, the shape becomes W: lows 0.01 and about 0.01007760, neckline about 0.01052632. Now (1/99.23−1/100)/(1/95−1/100) ≈ 14.7435%, which passes. The reciprocal changes relative lengths. Record price orientation along with the denominator; corresponding M/W names do not ensure identical candidate sets.

A shape detector, a breakout observation and a trading strategy are separate rules. State each one, including inputs, recognition time and an evaluation horizon. A strategy additionally needs entry, exit, size, costs and fill assumptions. The letter M or W alone does not supply them.

The neckline — the most important line of M and W patterns

In this model, M’s neckline is the selected intermediate minimum; W’s is the selected intermediate maximum. Require a close strictly below it for M or strictly above it for W, on an unchanged timeframe. A close exactly on the level fails our condition. This makes classification reproducible without guaranteeing the next close or an actionable profitable trade.

A double-pattern neckline uses an intermediate point as a horizontal price level. A head-and-shoulders construction can connect two intervening lows or highs and may slope. This is a difference in geometry, not a ranking of exhaustion or predictive strength. Record which point-selection and line rule is being used.

For M with neckline 1.1120, a low of 1.1115 followed by close 1.1123 does not meet our close rule; a later close 1.1118 does. For W at 1.0880, a high of 1.0886 and close 1.0878 fails; a later close 1.0882 meets it. An intrabar crossing and a close beyond the line are different events and should not be interchanged after the outcome.

Suppose the next closes are 1.1130 for M and 1.0876 for W, both back inside. Under an own definition of failure as a close returning inside within three following bars, both breakouts fail. Keep the earlier recorded completion: a later failure is an outcome, not permission to erase it. This failure rule is separate from the scanner’s named status logic.

What an activity field does and does not establish

If a test adds an activity filter, state the field, instrument, provider, comparison period and threshold. A higher bar does not by itself prove that banks confirmed the shape or that buyers or sellers changed intentions. It changes the candidate set under your rule, so evaluate accepted and rejected cases instead of assign a universal volume pattern to every reversal.

For MT5, Price Data defines tick volume as received ticks during bar formation; deal volume is a separate field and may be unavailable for OTC instruments. These quantities are not global FX notional turnover. Check actual data and settings before treating an activity comparison as evidence for a strategy.

The Forex Volumes help page describes its measure as price changes, while Price Data specifies received ticks. Preserve the scope of the supplied field and implementation rather than assume every platform’s volume counts the same events. No automatic global-volume inference or universal “twice the average” threshold follows. Results from a particular equity sample require their own rules and cannot establish an FX success probability.

Entry rules and the measured move

A full-height projection is a geometric reference, not a probability or time forecast. Our M/W points project 1.1040 and 1.0960 from the necklines; neither implies an 80-pip gain from any possible fill. A plan may test entry at a later available price or require a retest, with different observations and execution assumptions.

A retest is a return toward the crossed boundary. Price may move away again, return inside or never revisit it. Requiring a retest selects a different set than requiring the first close beyond the line; it does not automatically give a tighter stop or a safer trade. Define its recognition, expiry and outcome horizon, and do not relabel a recorded failure as an ordinary retest afterward.

Measure H from the more extreme point to the neckline under the chosen convention, then subtract H for M or add it for W. Stop, target, size and order type need separate rules; the shape does not specify the only valid stop or a compulsory buffer. An ordinary stop may fill beyond its planned level, so a planned distance is not a maximum possible loss.

A split exit or trailing stop is another rule to evaluate. Scaling out need not always reduce result variance; it depends on the complete distribution, sizing and execution. Nor does closing part of a position universally halve its best result. Record each actual fill and cost rather than infer a known outcome from the drawn projection.

False breakouts — when the pattern fails

A close beyond the neckline can be followed by a return inside. Define that event and its observation window before counting it. A later unfavorable result does not invalidate the fact that the earlier close met a chosen condition. Keep all cases, including unresolved observations, under the same rule.

Time of day, trend context, activity and H/ATR may be tested as additional conditions. Their names do not diagnose why an individual break failed or establish a universal reliability ranking. Record provider and timeframe, all tested variants and costs, then evaluate a fixed version on unused data.

Practical takeaway: Define the two extremes, intervening trough or peak, tolerance and breakout rule before identifying the pattern.

There is no compulsory combined H4/D1, activity and one-ATR defense that guarantees reliability. Choose conditions explicitly and measure how they change count, total and average results, execution and risk. A small pattern, a wider stop or a missed retest may lead to skipping a trade under a plan, but no candle letter settles that decision for every account.

Errors that change recognition and evaluation

Use these checks to preserve the rule and the information available at the time. They address reproducibility, not a claim that one timeframe or entry style is always best.

• Missing the tolerance denominator. Nominal price and pattern height are not interchangeable.
• Omitting an intermediate extreme. The full-interval minimum and first local trough can give different heights.
• Using later information earlier. Record recognition and completion time separately from plotted pivot bars.
• Changing failure after the outcome. Retain the chosen horizon and the prior completion event.
• Mixing projection with a fill. Distance from the neckline differs from distance from the executed entry; include costs and ordinary-stop slippage.

What to do when you close this page

A fixed rule can state extrema, similarity, context and completion explicitly. The examples do not reproduce a particular scanner’s detector, but they show why the denominator, orientation and recognition times belong in the record. Test the stated rule rather than fit a completed shape to a later move.

1. Change the second W low. Use 1.0813 with 1.0800 and neckline 1.0880. Check why 16.25% fails the illustrative 15% height rule.
2. Check reciprocal orientation. Recalculate the 100/99.23/95 model and its reciprocal. The same path fails at 15.4% but passes at about 14.7435%.
3. Record completion and outcome separately. Apply the strict close rule and three-bar return-inside definition to the supplied M/W sequence. Keep both completion and subsequent failure; do not require a fixed number of examples.

Related reading: head and shoulders — trend reversal pattern for comparison with the three-peak structure built on similar logic; how to draw support and resistance — without this skill you will not place a useful neckline; most important candlestick patterns — the full survey of patterns that complements the M and W structures.