Head and Shoulders — and What 'Reliable Pattern' Actually Means

Drawing it properly, which most people do not

1. 1It must come after an advance. A head-and-shoulders is a reversal pattern, so it needs something to reverse. Without a prior uptrend it is three bumps.
2. 2The head must be the highest peak and the shoulders roughly comparable in height. Wildly uneven shoulders usually mean you are drawing a pattern onto noise.
3. 3The neckline joins the two troughs, and it can slope. A neckline sloping downward is the more bearish version, because the second trough already broke the first.
4. 4Nothing happens until price closes through the neckline. This is the rule that separates the pattern from the drawing, and it is the one skipped most often.

Where the price target comes from, and what it is worth

Be clear about the status of this number. There is no mechanism connecting the height of a pattern to the size of the move that follows it — the rule is a heuristic that got repeated until it looked like a law. Measured targets are reached often enough to be useful for scaling and not often enough to be traded as a promise.

The practical use is in the ratio it gives you before entry. If the target is 8% away and your stop is 5% away, that is a different trade from one where the target is 20% away for the same stop — and you can compute that when the pattern completes rather than afterwards. See risk to reward.

The throwback, and why it looks like a failure

This is worth knowing in advance because it changes what you do rather than what you think. Two options, both defensible: enter on the break and set the stop above the right shoulder rather than just above the neckline, or wait for the throwback and accept that you will miss the patterns that never give you one.

What "reliable" means once someone counts

Thomas Bulkowski's contribution was not a new pattern. It was counting thousands of instances of the existing ones and publishing what happened, including the failures.

That work reframed the whole subject. Before it, chart pattern books showed examples that worked; the Encyclopedia of Chart Patterns reports break-even failure rates, average moves, and how often the measured target is actually reached — and the numbers are consistently less impressive than the textbook cases suggest.

• Failure is normal, not exceptional. Every pattern has a meaningful rate at which it breaks the wrong way or goes nowhere. A method that assumes otherwise is not a method.
• Measured targets are often missed. The projection is a scale, and a plan that only works if the full target is reached is a plan with a hidden assumption in it.
• Context changes the numbers. The same pattern behaves differently depending on the trend it appears in and where in a range it forms — which is why the figure below matters more than the shape.
• Published statistics are period-specific. They were measured on particular markets over particular decades. Use the ordering — which patterns behave better — rather than the decimal places.

How to use a pattern without being used by it

1. 1Require the prior trend. No advance, no reversal pattern. This single filter removes most of the false ones you would otherwise draw.
2. 2Wait for the close through the neckline. An intraday poke is not a completion, and it is where most of the disappointment comes from.
3. 3Compute the target-to-stop ratio before entry. If it is not worth taking on those numbers, the quality of the pattern is irrelevant.
4. 4Count your own. Twenty patterns of one type, recorded in R multiples, tell you more about how it behaves in your market and timeframe than any published table.