How often does a reversal pattern reverse anything?

You now know three reversal shapes and the behaviour behind each. This lesson asks the question that pattern books tend to skip: how often do they actually precede a reversal? The truthful answer is uncomfortable, and understanding why it's uncomfortable is more valuable than any number.

Why failures are structurally common

Three reasons, all mechanical:

Trends persist longer than shapes suggest. A "double top" inside a strong multi-year uptrend very often turns out to be a consolidation that the price later exceeds. Nothing about two similar peaks overrides the larger context in which they sit.

The visible level attracts deliberate probing. A level everyone can see is a level where stop orders cluster. Other participants know that. Prices routinely trade just past obvious levels, triggering those orders, before returning — sometimes because participants are deliberately probing for that liquidity, often just because that is where the resting orders happened to be. Either way it invalidates the neat shape and leaves the people who acted on it offside.

Confirmation arrives late by construction. As Lesson 1 showed, a reversal is only confirmed after both the failed rally and the broken low. By then a good deal of the move has already happened, so even a "successful" pattern often delivers its information after the fact.

A worked failure

Illustrative and rounded: a stock forms two peaks at $78.00 and $77.60 with a trough at $70.20 — a clean double top by anyone's definition. It breaks below $70 and reaches $67. Six weeks later it's back at $79, and four months after that it's at $95.

The shape completed. The break of the trough happened. And it reversed nothing — it marked a three-month pause inside a continuing advance. The chart of that episode, viewed today, doesn't look like a double top at all; it looks like an obvious dip. The pattern didn't change. Your view of it did, because you can now see what followed.

What the research actually says

Being precise here matters more than being reassuring:

• Some careful academic work has found measurable but small information content in automated pattern definitions on US equities (Lo, Mamaysky and Wang, 2000, is the standard reference).
• Other work — including earlier studies like Levy (1971) and various later replications — found effects that were weak, unstable across periods, or absent out of sample.
• Effects that survive tend to shrink or disappear once realistic transaction costs are applied, and they vary by market, era and liquidity.

So: not zero, not reliable, not settled. Any source quoting you a crisp "head and shoulders works 83% of the time" has quietly made a dozen definitional choices you can't inspect — starting with what counts as a head and shoulders.

The useful stance

Reversal patterns are best used as descriptions that generate questions: demand thinned here; the level that was defended stopped being defended; participation fell across these peaks. Those statements are checkable. "This is a reversal" is not, until later.

Try it now

1. Cover the schematic above from the marked broken trough rightwards and write down, in one sentence, what you would have said about it there. Then reveal the rest. Keep the sentence; it is the most useful thing on this page.
2. Now count rather than admire. In the five years below, scan for reversal shapes — double tops, double bottoms, head-and-shoulders — find at least five candidates, and for each record whether the prior trend actually turned over the following three months. Don't adjust the definition afterwards to improve the score; that's the exact habit Unit 4 exists to break.

3. Your list will contain failures. Pick the most convincing-looking shape that failed and write one sentence explaining what you'd have believed about it at the right edge.