Double top and double bottom — classic reversal patterns

A double top or bottom can organise a proposed reversal, but a shape is not a complete strategy. Specify the points, confirmation time, execution and costs. This article uses a fictional EUR/USD double top to compare a chart projection with actual quote-side fills. The numbers are arithmetic exercises, not a forecast or measured success rate.

What exactly is a double top and a double bottom?

A double top is an M-like arrangement after a rise: two nearby highs separated by a trough. A double bottom is the mirrored W-like arrangement after a decline. The price shape does not identify who tried to buy or sell or establish why the turning points occurred.

The element between the peaks or troughs matters most. With a double top it is the trough separating the two attempts; with a double bottom, the local high between the lows. We call that level the confirmation line, or neckline. Until price breaks through it, the pattern does not exist yet — you only have two turning points that may still be a pause inside a trend.

Why doesn't the second peak have to match the first exactly?

Nearby peaks need not be identical, but tolerance needs a stated value and denominator. A lower second high does not by itself establish greater predictive strength. Define whether a higher second peak is allowed by the chosen tolerance; it is not automatically proof that the rally will continue.

Record an explicit peak-tolerance rule before testing. The hypothetical example below has a 10-pip peak difference and an 80-pip pattern height, so its difference is 12.5% of height. It does not pass a 3%-of-height filter; a variant accepting it must state a tolerance of at least 12.5%. Tolerance and invalidation are separate rules, not universal definitions.

How do you confirm the pattern and measure the target?

Our model requires a Bid candle close below the neckline for the double top, mirrored for a bottom. This is a chosen confirmation filter, not evidence that most false signals disappear. A wick crossing and a close beyond are different events. Keep the rule unchanged within a test; the support and resistance guide gives context.

IG’s example measures height from the neckline to the highest peak and projects it from the neckline in the breakout direction. It is a reference, not a promised fill or market limit. Stops beyond a peak and retest entries are extra conditions to define and test. Keep cases with no retest and failed retests; a more attractive historical entry price cannot be assigned if it was never available or executed.

A hypothetical worked example, step by step

Here is the logic on daily EUR/USD — hypothetical only. After weeks of gains the pair stalls at 1.0900, pulls back to 1.0820, then returns and forms a second peak at 1.0890, a touch lower. The trough at 1.0820 is the confirmation line, and the pattern stands 80 pips tall, measured from the higher peak at 1.0900 down to that line.

In a separate chronology for these hypothetical Bid levels, define a pivot as strictly above or below two bars on each side, rejecting ties. The first high at bar 2 becomes known after bar 4, neckline low at bar 5 after bar 7, and second high at bar 8 after bar 10. Assume neighbouring values satisfy those pivot tests. Bar 10 closes at 1.0830, above the 1.0820 neckline; bar 11 closes at 1.0810, below it. Only then are the model’s pivots and close confirmation available. No entry at bar 8 can be justified by information from bar 11.

Assume a later actual short sale of EUR 10,000 at Bid 1.0808, with simultaneous Ask 1.0810, on a USD account. This is distinct from bar 11’s historical Close 1.0810. At projected Bid 1.0740 and a modelled 2-pip spread, Ask is 1.0742. Closing the short at that Ask earns USD 66 before fees, not the USD 68 obtained by incorrectly using Bid for the buyback. In MT5’s OTC convention, short SL/TP checks use Ask; reaching a Bid chart target alone does not establish that an Ask-based TP at that same numeric price has triggered.

The stop is an assumption, not a universal invalidation level or a guaranteed fill. Spread is already in the Bid/Ask prices and is not deducted again; financing, conversion and other fees are omitted. Net potential gain/loss is 64/99 ≈ 0.6465. A ratio below one does not alone disprove a strategy, just as one above two does not establish an advantage: probabilities, the full outcome distribution, costs and drawdowns also matter.

Fix the height convention separately from peak tolerance. For our two highs 1.0900 and 1.0890 and neckline 1.0820, highest-peak height is 80 pips and target 1.0740. Using their mean 1.0895 instead gives 75 pips and target 1.0745; using the lower peak gives 70 pips and target 1.0750. Those alternatives are our comparisons, not IG’s highest-peak convention. A 10-pip change of chosen peak moves the target by 10 pips. Do not select a convention after seeing which later price was reached, or move this neckline-based projection with a delayed entry.

What is the honest success rate of this pattern?

Historical stock-pattern statistics cannot be treated as this FX setup’s success rate without the original sample, pattern definition, failure metric and execution assumptions. No verified EUR/USD hit rate is provided here. FX also can have weekend gaps and intraday quote gaps; continuous weekday access does not guarantee a price at every level. Test the complete rule on the chosen provider data and retain failed cases.

A prior rise or decline is part of the chosen reversal-pattern definition, but a trend filter or activity condition does not automatically improve results. For OTC currency charts, specify what any volume field measures before treating it as market-wide participation. Compare a changed rule separately with the trend-following systems context; the technical analysis course covers related tools without supplying a success rate for this setup.

How is it different from head and shoulders?

A double top compares two nearby highs; a head-and-shoulders construction has three main peaks with the middle head higher than the shoulders. Its neckline uses the two intervening troughs and may slope. The head and shoulders pattern guide explains that geometry. The shapes alone do not establish relative frequency, ease or predictive reliability, nor require identical confirmation, stop or execution rules.

The most common mistakes

Common test errors include using future-confirmed pivots for an earlier entry, changing wick and close rules after outcomes, failing the declared tolerance and treating a quote touch as proof of execution. A lower timeframe still permits defined patterns; it does not erase their meaning by itself. Keep projected targets, triggered orders and actual fills separate and assess net outcomes without transferring unverified stock-pattern statistics to Forex.

What to do tomorrow

1. Choose a dated series and freeze pivot, trend and peak-tolerance rules. Record when every required point becomes known.
2. Record the neckline and chosen wick or close confirmation with a timestamp, then preserve rejected and failed cases.
3. Specify the height convention and project from the neckline. Calculate net gain/loss from actual or explicitly assumed quote-side fills rather than reject every ratio below one.
4. Record the tolerance value, denominator and boundary inclusion in the journal. The example’s 10-pip difference is 12.5% of its 80-pip height and does not pass a 3% filter.
5. Evaluate unchanged rules on later data, including costs, missed retests and worse fills. A dozen demo signals or a repeatable short sample does not establish future profitability or live readiness.