Do candlestick patterns predict crypto prices?
Not after costs, in our test. Of the seven candlestick patterns a 2026 study singled out as predictive on hourly crypto data, only two kept their sign on 2022–2026 perpetual futures, at a fraction of the reported size, and both effects disappear when the trade is entered one minute after the candle closes. No pattern covered a typical altcoin trading cost.
The claim
Moser and Brauneis tested 55 classic reversal patterns from the TA-Lib library on hourly bars of almost 2,000 spot trading pairs across 36 exchanges, July 2018 to January 2022, with a trend condition and a stepwise test built for many simultaneous hypotheses. Seven patterns stand out — bullish Harami, Hikkake, Doji Star and Inverted Hammer; bearish Harami, Doji Star and Hanging Man. The Hanging Man was followed by an average 0.56% drop in the next hour. Trading costs were not included.
What we tested
We applied the same pattern definitions, library version and trend condition to hourly perpetual futures on BTC, ETH and several hundred altcoins, including delisted contracts, 2020 to 2026, and judged the seven named patterns on the years after the paper’s sample. Each pattern was compared with the same logic run on bars with randomly flipped signs, which keeps volatility and the trend filter but removes real predictive content. We also measured returns from a price one minute after the candle closed — the price a trader could actually get.
Result
On 2022–2026 data only the Hanging Man and the Inverted Hammer kept the paper’s sign, and only over the first hour measured from the candle’s own close: 0.03% and 0.09% in the predicted direction, against 0.56% and 0.42% in the paper. From one minute after the close, the Hanging Man effect is zero and most of the Inverted Hammer’s is explained by where the candle closed within its range. Even inside the paper’s own period, on futures, the effects were 2 to 35 times smaller. The best break-even cost across patterns and horizons was 0.032% per side, below the cost of trading altcoin perpetuals.
Next-hour move after the pattern, in the direction the paper predicts
| Pattern | Paper 2018–22 | Ours 2022–26 | Entry 1 min later |
|---|---|---|---|
| Hanging Man (bearish) | 0.56% | 0.03% | ≈ 0 |
| Inverted Hammer (bullish) | 0.42% | 0.09% | mostly bounce |
| Five other named patterns | 0.04–0.34% | none | — |
“Ours” is measured from the candle’s close, “entry 1 min later” from the first price a trader could get. “Mostly bounce”: explained by where the candle closed within its range. “None”: indistinguishable from chance.
What it means
Much of what makes candlestick patterns look predictive on hourly data is mechanical. A candle that closes near its high tends to tick down next and one that closes near its low tends to tick up, and the effect is about three times stronger on thinly traded pairs. That is the bid-ask bounce of the last trade, and it is gone by the time an order can be filled.
What would change the verdict
The verdict could change with maker execution at the candle’s close, on illiquid spot pairs where the spread itself is the effect, or with newer versions of the pattern library that redefine several patterns.
This note re-tests a claim made by the cited authors on our own crypto market data; the summary of their method is ours. Tested · published · last reviewed . Research, not investment advice.