I'm testing different systematic trading strategies to see the impact they have on the portfolio's performance metrics (volatility, returns, Sharpe) and wanting to do this via simulation. I'm wanting to generate the asset returns and run it through the backtests and compare the differences in said metrics. To generate the asset returns, I've tried the following:
- Circular Block Bootstrapping each asset class individually
- Circular Block Bootstrapping all asset returns together (maintains covariance exhibited)
- Randomly generated returns picked from multivariate normal distribution
While these return-generating methods do provide some insight, they have the following limitations:
- Doesn't take into account covariance between each asset class
- Not sure of the validity of this method. Each asset class has a different ideal block length to preserve their respective autocorrelation; only one block length can be used here to represent all asset classes
- Doesn't take into account autocorrelation within each asset class
Are there any methods out there that can randomly generate asset returns which maintain autocorrelation and covariance?