Crypto currencies Trading Strategy based on Markov Switching GARCH
Data from https://www.cryptodatadownload.com/data/gemini/
- Trade on a group of crypto currencies
- Ensemble technical indicators in Random Forest to predict price movements
- Portfolio Allocation based on volatility predictions
- Set trading frequency as hourly
- Used all asset data for model training
- Modeled each asset's risk profile separately
Use Cusum filter(Lam, K. and H. Yam (1997): “CUSUM techniques for technical trading in financial markets.” Financial Engineering and the Japanese Markets, Vol. 4, pp. 257–274) to sample the data points that observed significant price movements
Train the model with the selected samples.
- Select from a bunch of technical indicators that illustrate either momentum or volatility
- Average Directional Index
- Commodity Channel Index
- Moving Average Convergence/Divergence
- Price Momentum
- Relative Strength Index
- William's % R
- On Balance Volume
- Slope of Linear Regression
- Return Volatility
- Use random forest as a regressor to predict the direction of movements
- Relatively high frequency, used non-linear model for more convexity
The general Markov-Switching GARCH specification can be expressed as: (Ardia et al. 2017 https://doi.org/10.1016/j.ijforecast.2018.05.004.)
where
An illustration of MSGARCH on SPY: the black line is the probability of being in the high volatility regime
- Detect Volatility Regime for each instrument
- Predict future volatility as the input of allocation model to replace realized volatility
-
Rebalance weekly, based on forecast volatility, following minimum variance principle across assets
-
Use forecast instead of historical volatilities to construct covariance matrix
where
See the notebook here
[1] Lam, K. and H. Yam (1997): “CUSUM techniques for technical trading in financial markets.” Financial Engineering and the Japanese Markets, Vol. 4, pp. 257–274
[2] David Ardia, Keven Bluteau, Kris Boudt, Leopoldo Catania, "Forecasting risk with Markov-switching GARCH models:A large-scale performance study", International Journal of Forecasting, Volume 34, Issue 4
[3] APA. Lopez de Prado, M. (2018). Advances in financial machine learning, 38-40
[4] Jurczenko, Emmanuel, ed. Risk-based and factor investing. Elsevier, 2015. 10-12