Section pillar · 259 pages
Every formation written as a rule, then run over 1.8 million bars.
Textbook technical analysis tells you what a pattern means. This section tells you how often it occurred, what followed it, and whether the difference is large enough to trade. Where the answer is no, the page says no.
So you can tell textbook consensus from our own measurement without inferring it from tone.
Uncontested and verifiable outside this site.
What the textbooks assert, reported as their claim.
How the rule was defined and detected.
Our own output, with the sample size attached.
Too few occurrences to say anything. Said plainly.
Candlestick patterns · 107 formations
107 formations written as conditions over open, high, low and close, each with the thresholds it uses and the conditions that invalidate it. Search them, filter by direction or candle count, and open any one for its rule and its verdict.
Chart patterns · 20 formations
Head and shoulders, double tops, wedges, flags and triangles, each drawn from machine-checkable conditions: what has to be true, what the conventional target is, and what invalidates it.
Not measured here
The chart formations carry geometry and the traditional claim, not our own forward-return study. That work is not done yet, and the pages say so rather than implying a verdict.
Log in and run any formation through the sizing and performance tools, with the sample size and the verdict attached.
Technical analysis is four fairly different activities that share a name. Only the first is published here, and the reason is not that the others are worthless — it is that the first is the one that can be written as a rule and checked.
| Kind | Covered here |
|---|---|
| Price formations — candlesticks, chart shapes | Yes. 125 of them, each as an executable condition with published thresholds |
| Indicators — moving averages, RSI, MACD | Only inside the [strategies](/strategies) section, as complete rules rather than as indicators |
| Market structure — support, resistance, trend lines | Not yet. Drawing them involves choices that two analysts make differently, so a rule would be testing the choice rather than the structure |
| Discretionary reading — "the chart looks weak" | No. Not because it is wrong, but because it cannot be stated precisely enough to be wrong |
The boundary is whether a claim can be falsified. A formation written as a condition on open, high, low and close either fires on a bar or it does not, and two implementations can be checked against each other. A trend line drawn by eye cannot be, which is why it is absent rather than dismissed.
Almost every disagreement about technical analysis comes from treating these as one thing.
Every formation page here separates them explicitly, labelled as the rule, the traditional claim, and — when the study is republished — the measured result. A page that blends them reads more smoothly and tells you less.
Worth knowing
The most useful section of a formation page is usually the one describing how the detection is wrong: which other formation it is geometrically identical to, and what separates them. That is true whether or not either has a measurable edge.
A signal is a recommendation you cannot audit. A rule with published thresholds is a claim you can run yourself, on your own data, and disagree with.
The practical difference shows up when something stops working. A signal service that degrades gives you no way to tell whether the market changed or the method was never sound. A rule with a stated threshold, a benchmark and a sample size gives you somewhere to look — and, when the answer is that the run was flawed, a reason to withdraw it rather than defend it.