There are countless articles about crypto candlestick patterns and their ability to predict market movement. Is there some truth or is this a pure myth? Let’s use the data to bust or confirm it.
Every crypto or stock market trader is familiar with the candlestick chart and uses it regularly. The chart represents price movement as a sequence of candles. There are many different types of candles, that can all tell a story about current market sentiment. Some candles even have their own names.
Let’s use the power of data science to compare the patterns with the actual market movement in the past.
Crypto candlestick patterns vs stock candlestick patterns
There is one big difference between the stock and crypto market. The stock market has opening hours, while the crypto market is always open, which results in a slightly different candle sequence. The closing price in the previous candle is always the same as the open price in the current candle. Therefore some popular multi candlestick patterns are not possible in crypto world.
Popular single candlestick patterns
My analysis will focus on single candlestick patterns with 1-hour candle timeframe. In other words, 1 candle will represent 1 hour of trading activity. For additional information about candlestick patterns head to the Getting started with candlesticks blog.

*Market returns as advertised in internet articles like this.
Baseline candlestick pattern analysis
About the data
Source: Kaggle
Timeframe: 2018-01-01 to 2021-09-21
Cryptocurrencies: Bitcoin & Ethereum
Let’s plot the baseline market returns chart for the whole dataset.

How to read the chart
- The upper chart shows mean log-returns comparing today and day x in the future.
- The lower chart shows the probability of positive market returns comparing today and day x in the future.
- Grey bars indicate statistically insignificant results (z-test).
Observations
- Total mean log-returns are slightly positive, indicating the overall positive trend in the crypto market.
- Positive returns probability is slightly above 50% which is also a positive trend indicator.
Let’s replot the chart for the special cases where the current market is in the downtrend / uptrend.

Observations
- Few statistically significant results when the market is in a downtrend / uptrend.
- Slightly above average mean log-returns when the market is in the uptrend.
Bullish marubozu analysis
Marubozu has a long real body with very small upper & lower shadows. A green candle is representing bullish marubozu, while a red candle is showing a bearish marubozu.

Advertised market movement: positive

Observations
- Mean log-returns are far above the baseline returns and positive probability is exceding 70%.
- Results are even better in the uptrend market.
- There are no statistically significant results in the first three days after the bullish marubozu pattern.
- Actual and advertised market returns are consistent.
Bearish marubozu analysis

Advertised market movement: negative

Observations
- Most results are statistically insignificant.
- Statistically relevant results actually show positive market returns, especially in a downtrend.
- Actual and advertised market returns are inconsistent.
Hammer analysis
The pattern has a long lower shadow and a small upper shadow. A green candle is representing the Hammer pattern, while a red candle is showing the Hanging man.

Advertised market movement: positive

Observations
- Baseline returns are mostly exceded in the uptrend.
- Downtrend returns show limited statistically significant data.
- Actual and advertised market returns are partly consistent.
Hanging man analysis

Advertised market movement: negative

Observations
- Few statistically significant results.
- Slightly above average positive probability 1 week after the pattern.
- Actual market returns are indecisive.
Inverted hammer analysis
The pattern has a long upper shadow and a small lower shadow. A green candle is representing Inverted hammer pattern, while a red candle is showing the Shooting star.

Advertised market movement: positive

Observations
- Baseline returns are exceeded in the downtrend with the probability of positive returns at 65% on day 6.
- Uptrend returns are indecisive.
- Actual and advertised market returns are partly consistent.
Shooting star analysis

Advertised market movement: negative

Observations
- Few statistically significant results.
- Above-average positive probability on day 6 after the pattern.
- Actual and advertised market returns are inconsistent.
Spinning top analysis
The pattern has long upper and lower shadows. The Spinning top has slightly bigger real body, while Doji’s real body is very small.

Advertised market movement: indecisive

Observations
- Negative returns are observed in total and downtrend market.
- Statistically insignificant log-returns in the uptrend.
- Actual and advertised market returns are inconsistent.
Doji analysis

Advertised market movement: indecisive

Observations
- Negative returns are observed in total and downtrend markets, however, they are closer to zero when compared to the Spinning top.
- Statistically insignificant log-returns in the uptrend.
- Actual and advertised market returns are inconsistent.
Conclusion
Crypto and stock market candle patterns can’t be directly compared. Some multi candlestick patterns can’t even be generated in the crypto market due to the fact that the market is always open.

The conclusions above are based on mean values with relatively high standard deviations. Pattern thresholds are also finetuned to maximize the statistical significance of data. Furthermore, data is only based on Bitcoin and Ethereum. For more details about the analysis head to my Kaggle notebook Crypto Candle Analysis.
Disclaimer: I am not a financial expert; the analysis above is the result of my own curiosity. Before making any investment, you should always do your own research.