Forecasting bitcoin returns with Prophet in Python— Part II
Introducing additional data to the model
In the previous article on forecasting bitcoin returns with Prophet I have discussed how I used an additive and a multiplicative model of which the coefficients were estimated with Prophet in Python to forecast bitcoin daily returns based on nearly 9 years of daily return data. In this article I introduce some extra parameters to the model and infer a price prediction from the daily return predictions.
Introducing Stock-to-Flow ratio as an additional independent variable
Bitcoin’s scarcity is controlled by the Bitcoin protocol. As the number of bitcoins that is created (via block rewards to miners) and added to the supply, halves every 210.000 blocks (nearly every 4 years), the scarcity increases over time.
A common measure for scarcity is stock-to-flow. If you like to learn more about this measure, I highly recommend reading Saifedean Ammous or to have a look at the work of PlanB. The concept of stock-to-flow ratio is as straight forward as the name of the metric suggests. It’s the ratio of stock (or supply) over the newly minted coins (or mined commodity) in a year.
Model and estimation of coefficients
The model is as follows:
Return(t) = trend(t) + X(t)*beta + noise,