What Happens if you Day Trade Crypto with Facebook Prophet?
Results of backtesting a Bitcoin intraday trading strategy using Facebook’s machine learning model
In the past, I’ve experimented with various forms of machine learning models in order to develop a successful cryptocurrency trading strategy. These various strategies would utilize classification models, time-series models, and sentiment analysis. The results of these strategies would range from mediocre to “good but requires more testing”. I’ve never settled on a specific strategy and hope to one day move on from the testing phase to “actually implementing it with real money” phase.
In all of these different strategies, I have yet to change the frequency of the historical data tested. The frequency of data I have always used and backtested with was the daily historical price data. That frequency for price data is not bad at all but it is limited, especially with the around the clock trading availability of the cryptocurrency market.
So in order to continue my testing, I’ll need to know what it is like to day trade crypto with machine learning. Specifically, I’ll be using a time series machine learning model that I have found to be easy to use and implement — Facebook Prophet. Feel free to code along with me or just continue…
