This page states how the numbers on this site are produced, what they assume, and what they cannot tell you. If a figure cannot be reproduced from a stated method, it should not be here.
Every claim on this site carries one of these five labels, so you can tell textbook consensus from our own measurement without inferring it from tone.
A fact that is uncontested and verifiable outside this site, such as a contract size or a formula's definition.
What technical-analysis textbooks assert about a pattern, reported as their claim, not as something we have verified.
How a rule was defined and detected: the exact condition over open, high, low and close.
Our own output from running that rule against the dataset, always shown with its sample size.
Too few occurrences to say anything reliable, or not measured on this dataset yet. Said plainly rather than implied.
Pattern research is generated at build time from stored rule definitions run over a canonical price feed. Every published figure carries the engine version and the data vintage it came from, so a number can always be traced to the run that produced it.
Detection is mechanical. A formation is a condition over open, high, low and close, with a trend filter applied before an occurrence is counted. Counts below the sample floor are reported as unresolved rather than dressed up as findings.
A quantitative trading research site with tools attached. Most trading content explains a method and asserts that it works. This site defines the method precisely enough to run, tests it against historical data, and reports what came back, including when the answer is that nothing did.
The calculators exist because a trade has to be sized before any of the research matters. The research exists because a number without context is just a number.
Corrections are welcome and get published. If a figure on this site is wrong, tell us which page and which number.
Risk disclosure
Everything on this site is educational material, not investment advice, and nothing here is a recommendation to buy or sell any instrument. Leveraged trading can lose more than the amount deposited. Past performance and historical research do not indicate future results. You are responsible for your own decisions and for verifying any figure before acting on it.
What is published here is a corpus: 125 formations — 107 candlestick and 18 chart — each written as an executable rule with its thresholds, its variants and the conditions that invalidate it. If you quote a threshold or a rule, cite the version, because changing a threshold changes every count derived from it.
Suggested citation
Custom Trading Platform (2026). Candlestick and chart formation definitions, with detection rules (version cfb08765de7c). https://customtradingplatform.com/candlestick-patterns
Licensed CC BY 4.0. The version string is the engine fingerprint that produced the definitions, and it is the same identifier the site uses to decide whether a measured run may be published — a figure computed on a different fingerprint is withheld rather than shown.
| Available now | Status |
|---|---|
| Formation rules and thresholds | Published, versioned cfb08765de7c |
| Strict and loose variants | Published |
| Invalidation conditions | Published |
| Engine-to-browser parity | 81,301 detections, 0 differing |
| Chart formation firing rates | Published, one symbol and timeframe, stated on the page |
| Hit rates, benchmarks, intervals | Withheld — the run was computed on a superseded engine and withdrawn |
No bulk download is offered. The definitions are published page by page, in readable form, rather than as an archive — the thresholds are all there, and each is attached to the explanation of what it decides.
The whole reference is also served as plain text, so a model or a script can read it in one request rather than crawling 181 pages: /llms-full.txt. It carries every formation with its definition, its detection thresholds, its strict and loose variants and the conditions that invalidate it, each beside its own URL — and the risk arithmetic, which is derivation rather than measurement.
It states in its opening lines that no hit rate, win rate or forward-return figure appears anywhere in it, because a file designed to be quoted verbatim is the worst place to put a number the site does not currently stand behind. A shorter index of every page is at /llms.txt.