Patterns divide into two families by what they claim. A continuation pattern claims the prior trend resumes: triangles, flags, pennants, rectangles and the cup-with-handle. A reversal pattern claims it ends: head and shoulders, double and triple tops and bottoms, rounding tops.
The head and shoulders is the canonical reversal and the only one worth learning to compute carefully, because everything else follows the same arithmetic. It is three peaks, the middle one highest, with the two troughs between them joined by a line called the neckline. The pattern is not a pattern until the price closes below that neckline, and the measured move is the depth of the pattern projected down from the break.
How is the head and shoulders target calculated?
Depth is where authors quietly disagree. Take the numbers from Worked Example 4: a left shoulder at ₹1,842.00, a head at ₹1,968.00, a right shoulder at ₹1,851.00, and neckline troughs at ₹1,704.00 and ₹1,668.00 eighteen sessions apart. The neckline slopes at minus 2.00 points per session. Under the head it sits at ₹1,686.00, so the depth is 1,968.00 minus 1,686.00, which is 282.00 points. Project that from the neckline where the break happened, at ₹1,652.00, and the target is ₹1,370.00. Project it from the neckline under the head instead, and the target is ₹1,404.00. Same pattern, same data, two published conventions, 34.00 points of difference, which is 2.48% of the target.
A book that gives you one target without saying which projection produced it has hidden a number the size of a decent quarter's return.
Advertisement
How do you trade a head and shoulders pattern?
The trade is what makes the target meaningful. Sell at the break close of ₹1,646.00, place the stop at ₹1,700.00 above the neckline and under the right shoulder, and take the lower target. Reward is 276.00 points, risk is 54.00 points, and the ratio is 5.1111 to one. The break-even hit rate is 54 divided by 330, or 16.3636%. At a 20% hit rate the expected outcome is plus 12.00 points per trade; at 30% it is plus 45.00.
Now the case the textbooks skip. In the failure branch of that same example the index falls to ₹1,598.00, which is 17.02% of the measured move, and then closes back at ₹1,712.00 and takes out the stop. Best open profit 48.00 points, realised loss 66.00 points, one full target's worth of reward foregone. The pattern was correctly identified by every published criterion. The direction was even right for a while.
And the trade lost money, because a pattern gives you a direction and a trade needs a distribution.
How reliable are chart patterns?
That is the honest way to hold all of these. A pattern is a hypothesis with an entry, a stop and a target, and it is worth something only if you know its hit rate on a sample that was not selected after the fact. Thomas Bulkowski's Encyclopedia of Chart Patterns is the serious attempt at that counting, and its numbers are far less flattering than the folklore. Even there, the identification criteria carry enough discretion that two analysts applying them to one chart will not always agree.
A rule two people cannot apply identically is a rule you cannot backtest.