Do candlestick patterns work? 22 patterns measured on four markets
Hammers, engulfing candles, morning stars: every trading course teaches them. We counted what actually happened in the five candles after each one, compared with an ordinary candle, on four markets.
Across 88 pattern–market tests (BTCUSDT, ETHUSDT, SOLUSDT, SPY, daily candles), 0 showed a statistically clear edge in the textbook direction and 0 worked in the opposite direction. 31 were measured and showed no edge; 57 appeared fewer than 20 times, too rarely to judge.
BTCUSDT: 1410 daily candles, 2022-12-11 to 2026-10-20.
| Pattern | Textbook says | Seen | Edge, 5 bars | t | Worked / base rate | On this market | |
|---|---|---|---|---|---|---|---|
| Spinning top | Indecision | 217 | -0.23% | -0.80 | 100% / 100% | No edge | |
| Doji | Indecision | 171 | -0.45% | -1.78 | 100% / 100% | No edge | |
| Bullish engulfing | Bullish | 70 | -0.12% | -0.16 | 51% / 52% | No edge | |
| Bearish engulfing | Bearish | 70 | +0.60% | 1.03 | 53% / 48% | No edge | |
| Bearish harami | Bearish | 55 | +0.95% | 1.81 | 58% / 48% | No edge | |
| Bullish harami | Bullish | 40 | -0.03% | -0.04 | 60% / 52% | No edge | |
| Tweezer top | Bearish | 25 | -0.21% | -0.21 | 48% / 48% | No edge | |
| Dark cloud cover | Bearish | 21 | +0.13% | 0.12 | 48% / 48% | No edge | |
| Hanging man | Bearish | 20 | -0.65% | -0.60 | 35% / 48% | No edge | |
| Tweezer bottom | Bullish | 18 | +0.95% | 0.49 | 61% / 52% | Too few | |
| Hammer | Bullish | 16 | +0.19% | 0.10 | 63% / 52% | Too few | |
| Piercing line | Bullish | 16 | +0.34% | 0.26 | 63% / 52% | Too few | |
| Shooting star | Bearish | 12 | -0.88% | -0.43 | 42% / 48% | Too few | |
| Morning star | Bullish | 8 | -2.01% | -0.90 | 50% / 52% | Too few | |
| Dragonfly doji | Bullish | 7 | +1.75% | 0.90 | 71% / 52% | Too few | |
| Evening star | Bearish | 6 | -1.71% | -1.03 | 50% / 48% | Too few | |
| Bullish marubozu | Bullish | 4 | -0.43% | -0.26 | 25% / 52% | Too few | |
| Three white soldiers | Bullish | 4 | +3.63% | 1.35 | 100% / 52% | Too few | |
| Inverted hammer | Bullish | 3 | +6.81% | 1.15 | 100% / 52% | Too few | |
| Gravestone doji | Bearish | 0 | — | — | — | Too few | |
| Bearish marubozu | Bearish | 0 | — | — | — | Too few | |
| Three black crows | Bearish | 0 | — | — | — | Too few |
Method
Patterns are detected with fixed rules (body and wick proportions plus the short-term trend before the candle), the same definitions used in the product and listed in the academy.
For each occurrence we measure the move over the next 5 candles in the direction the textbook predicts, and subtract the same measurement averaged over all candles. That difference is the edge; its t statistic says whether it is larger than chance.
"Worked" requires at least 20 occurrences and a clear edge (|t| ≥ 2) on at least two of the horizons 1, 3, 5 and 10 candles, in the same direction.
Limits — read before quoting
Daily candles only; no fees; patterns tested on their own, not combined with trend filters or levels — combinations are what the strategy studio is for. A pattern that shows no edge on its own can still be useful as a trigger inside a strategy; test it there.
Last recalculated: October 20, 2026 · Results may be quoted with a link (CC BY 4.0).