Premium Market Analysis, Trader Education, Software and Trading Strategies


Bitcoin Returns Are Not Serially Correlated

Bitcoin returns are not serially correlated as some traders think. More importantly, serial correlation and direction of the trend are not directly related.

I used bitcoin continuous BTC futures prices from Norgate data because of the well-defined trading hours (Sunday – Friday 5:00 p.m. – 4:00 p.m/ CT) as opposed to free data some traders use from the web that may not have well-defined or consistent timing of the closing price.

Below is a daily chart of continuous BTC prices with 50-day autocorrelation and Hurst exponent plots.

The autocorrelation chart shows that contrary to some interpretations, BTC has fallen when autocorrelation was positive (blue.) In fact, some strong rebounds have occurred during periods of negative autocorrelation (red.) This makes sense to me but going into more detail is beyond the scope of this brief article. However, it’s a common misconception that uptrends need positive serial correlation and the stock market since 2000 is a proof of this. Note that average autocorrelation is -0.04 meaning that on the average there is no serial correlation.

Furthermore, asset prices can uptrend or downtrend irrespectively of the sign of autocorrelation. What determines the direction of the uptrend are the relative changes, not the autocorrelation sign.

The Hurst exponent basically reveals the same pattern as the autocorrelation: values below 0.50 indicate mean-reversion (negative serial correlation) and above 0.50 momentum (positive serial correlation.) As it turns out, the average Hurst for 60-days lookback period is exactly 0.50 meaning this market has been on the average neither in mean-reversion nor in momentum. This is interesting and again discussing it further may require another article.

Finally, as a word of caution, descriptive statistics have little application to trading but are (possibly) useful for asset allocation and strategic investing. I indicated “possibly” because choice of lookback period may lead to significant variations in results, as anyone who has actually gone through the exercise already knows.

Disclaimer:  This article does not include a trade recommendation. Read the full disclaimer here.

Charting and backtesting program: Amibroker. Data provider: Norgate Data

If you found this article interesting, you may follow this blog via RSS or Email, or in Twitter.

15% off for blog readers and Twitter followers with coupon PAL15

Click here to subscribe

Premium Content