I am working on the comparison of different volatility timing/target strategies on portfolios starting from different conditions (data, asset classes, calculation of realized volatility, different exposure assumptions, etc...).

I was thinking that it would be interesting to see what happens if I simulate volatility exceptional behaviors throughout the data in the backtests, but I'm not sure how to implement that.

I tought of defining a stochastic process (maybe a point process?) ad hoc capable of that, but I don't know where to start to look for something theoretically solid.

Any advice to put me in the right direction is well appreciated.

P.S. If it can be useful, I am working on Python, mainly numpy, pandas, and the usual related packages.

  • 2
    $\begingroup$ If your objective is altering data to have random shocks, maybe introducing jumps instead of affecting historical volatility might produce better results. First of all, switch to returns from prices. GARCH filtering to get rid of autocorrelation and heteroskedasticity may help before bootstrapping. Then just arbitrarily set the parameter $\lambda$ of your Poisson distribution from scipy.stats, sample random numbers from it and let these numbers mess your returns. Assume some distribution for the magnitude of the jumps themselves, e.g. Normal. Finally, it's time for simulation. $\endgroup$
    – Lisa Ann
    Nov 1, 2019 at 23:17


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.