What it means to say a pattern worked

Saying that a pattern worked needs a definition, and this site fixes one before it counts anything: from the close that completes a figure, a hit is a close 2 ATR away in the textbook direction within 20 sessions, before any close 1 ATR the other way. The same definition is applied to every figure, on every one of 11 instruments, and it was written before any count was run.

The definition, in numbers

A hit is judged by one fixed definition, the same on every page. A figure is counted at the close that completes it (for a candle model, the close of its last candle). From that close C, and the ATR of that session (the average true range of the last 14 sessions), the textbook direction follows from the figure: down for head and shoulders, double top, descending triangle and bearish engulfing; up for double bottom, ascending triangle and bullish engulfing.

It is a hit if one of the next 20 sessions closes at least 2 ATR from C in that direction before any session has closed 1 ATR or more from C the other way. It is a miss if a close 1 ATR against comes first. If neither happens within the 20 sessions, the occurrence is undecided: it counts as not a hit and stays in the denominator, so waiting does not improve the share. Only closes decide, never highs or lows.

Why 2, 1 and 20

The three numbers were chosen once, from a textbook ratio of two to one between the move that counts and the move that cancels it, and written into the measuring tool before any figure was counted. They were not adjusted after the first results. That is the point of fixing them in advance: a definition tuned until a pattern looks good would measure the tuning, not the pattern.

The numbers are still a choice. Another site could call a hit 3 ATR within 30 sessions, and its shares would differ from these. Neither is wrong; they answer different questions. A reader comparing this site's shares with another site's should first compare the definitions.

Two real occurrences, one hit and one miss

The chart under the heading shows a bearish engulfing pair on Reliance Industries, completed on 22 January 2021 at a close of 945.89 with an ATR of 24.17. The target is 2 ATR, 48.34, below that close; the cancel level is 1 ATR, 24.17, above it. On 25 January a close came 2.07 ATR below C, past the target and without any close 1 ATR above first: a hit. The second chart is a bearish engulfing pair on Tata Consultancy Services completed on 28 February 2019 at 1,983.45, with an ATR of 55.94. On 15 March, after two weeks, a close stood 1.01 ATR above C: a miss.

Both were found by the same written rule, both are drawn on their real candles, and both are counted in the tables on the other pages. The two were chosen as a plain example of each outcome, not as proof that bearish engulfing pairs hit one time in three.

The yardstick: the same test from any session

A share means little alone, so every figure on this site is shown next to a yardstick. The same test, with the same 2 ATR, 1 ATR and 20 sessions, was run from every session of the same histories, once for each direction, with no figure needed. Over 26,729 sessions across the 11 instruments, a close 2 ATR up came before a close 1 ATR down in 38.0 percent of starts, and the mirror test down was passed in 27.8 percent.

Those two numbers say what an entry made on a coin-flip would have scored under the identical rule. A figure whose share is near its yardstick has shown no edge over starting on any day; one far above or below it has shown something, but only if the sample is large enough to trust, which on this site it usually is not.

Small samples and what is left out

Rare figures produce small counts: six head and shoulders, three double tops, two descending triangles in ten years across 11 instruments. With six occurrences, one result more or less moves the share by almost 17 points. Each table prints the count beside the share for exactly this reason, and the pages say so where it matters.

Two occurrences are left out of the shares because fewer than 20 sessions of history follow them, so the window could not be completed: one double bottom and one bearish engulfing pair. They are listed in the table as left out and are not counted as hits or misses. Instruments with no occurrence of a model are shown with zero; none is hidden.