Research · Candlesticks
The short answer
The ranking exists, but the gaps between the top entries are smaller than their confidence intervals.
That is a deliberate absence rather than an oversight. A reliability ranking requires measured outcomes, and the measured run for these formations has been withdrawn — it was computed on a superseded engine with the still-forming bar included.
Ranking without measurement would mean ordering 107 formations by tradition, which is what every other list does and what makes them unfalsifiable.
A ranking that clears all five will be short, and the gaps between the top entries will be smaller than their confidence intervals. That is the expected shape of an honest result here.
Most published rankings measure something else and label it reliability — usually frequency, or the ease of spotting the shape after the fact. Neither is predictive.
The one property that is both measurable today and genuinely useful is how the detection fails. Each formation page publishes it: the geometric neighbours it cannot be distinguished from without context, and the threshold that does the distinguishing.
Worth knowing
If a list ranks the doji as highly reliable, ask what it was compared against. A formation appearing on one bar in eight cannot carry much information whatever it precedes.
It will come from the re-run, with a benchmark, a corrected family of 125, a sample floor and per-market replication, and it will appear on the formation pages rather than as a bare list. Until then the candlestick patterns index carries every rule and every threshold, which is the part of the answer that does not depend on a measurement.
The word is used for at least four different properties, and a list ranking on one while implying another is the most common way these articles mislead.
| Reading | Measurable? |
|---|---|
| Precedes the claimed direction more often than chance | Yes — needs a benchmark and a correction |
| Produces a tradeable move after costs | Yes — a stricter test than the first |
| Appears often enough to matter | Yes, but it is frequency, not reliability |
| Is easy to spot on a chart | Not a property of the market at all |
Only the first two are claims about whether the formation carries information. A ranking that quietly uses the third or fourth is measuring convenience.
Any formation appearing in a rising market will be followed by a rise most of the time. That says nothing about the formation, and it is the single largest source of inflated reliability claims.
The comparison is the finding. Any figure quoted without the benchmark beside it is describing the market rather than the pattern, however precisely it is stated.
Search for the most reliable candlestick patterns and the top results will not agree, often dramatically. The disagreement is not a mystery and it is not evidence that one of them is right.
Two lists built on different definitions of the same word cannot be reconciled, and averaging them produces something less meaningful than either.