Using High-Frequency Entropy to Forecast Bitcoin's Daily Value at Risk

Area of Science:

• Quantitative Finance
• Computational Finance
• Econometrics

Background:

• Accurate Value at Risk (VaR) forecasting is crucial for financial risk management.
• Traditional models like GARCH have limitations in capturing the complexities of cryptocurrency returns.
• High-frequency data offers potential for improved risk assessment.

Purpose of the Study:

• To evaluate the efficacy of econometrical models in forecasting VaR for cryptocurrency returns.
• To explore the use of symbolic time series analysis (STSA) for estimating return distribution entropy.
• To compare the predictive power of entropy-based VaR forecasts against GARCH models.

Main Methods:

• Utilized high-frequency Bitcoin return data.
• Applied symbolic time series analysis (STSA) to estimate intraday log-return entropy.
• Employed Christoffersen's tests for Value at Risk (VaR) backtesting.

Main Results:

• Entropy of intraday returns demonstrated significant explanatory power for daily return quantiles.
• VaR forecasts derived from intraday return entropy significantly outperformed GARCH model forecasts.
• STSA effectively reduced high-resolution data to low-resolution data for entropy estimation.

Conclusions:

• Entropy derived from intraday return distributions is a superior predictor of Value at Risk (VaR) compared to GARCH models.
• Symbolic Time Series Analysis (STSA) is a viable method for feature extraction in financial time series.
• The findings suggest a novel approach for enhancing risk management strategies in volatile cryptocurrency markets.