Candlestick patterns with no statistics

[Shadow Trader]

Hammer, engulfing, morning star, shooting star. Books and videos present them with confident names and a handful of perfect examples.

What they rarely include is the hit rate. How often does an engulfing pattern on M5 EURUSD lead to a profitable move after costs? Studies that test candlestick patterns on large samples often find weak or inconsistent results, especially after accounting for spread.

That doesn't mean they're worthless. Patterns at significant levels, in the right context, may carry more weight than patterns in isolation. But "may" needs data.

A simple way to test: code the pattern exactly, run it across a year of data, and compare the forward returns to random entries. The difference, if any, is the real edge.

It's worth remembering how many patterns exist: dozens of named formations, each with variations. With so many possibilities, some will always appear to work in any given sample purely by chance. That's the multiple testing problem, and it quietly inflates the apparent success of pattern trading.

To give the idea some structure, here's what a fair candlestick test looks like in practice. Define the pattern in numbers: for an engulfing candle, the body must be at least 1.2 times the previous body and fully cover it. Define the context: only after three or more candles in the same direction, only during the first two hours of London. Define the trade: enter at the next open, stop beyond the pattern's extreme, target at 1.5 times the stop.

Then count. Run it over 18 months across two or three pairs, record every occurrence automatically, and compare the results with a control group of random entries taken under the same conditions at the same times.

The control group is the part that most tests skip, and it's the most revealing. A pattern that earns 0.1 pips more than random entries isn't a pattern, it's rounding error. A pattern that earns half a pip more, after costs, across hundreds of occurrences, might be worth a closer look.

Most of the famous formations come out indistinguishable from the control. A few, in the right context, do show something faint.

Has any candlestick pattern survived this kind of test, and what context was necessary for it to work?