So with the recent Covid induced market crash it becomes apparent that backtesting and walk forward analysis can be critically affected. Of course a 100% efficient/ robust trading strategy could deal with all eventualities the market could throw at it but I doubt such a system exists.
So that leaves a choice of trying to include the tsunami in the back testing period or test up to the date before the seismic shock and hope that the optimisation fits the market once reduced volatility returns.
Now that we have a few weeks of data since the major crash I’ve started looking at walk forward testing with a short range from the beginning of March on 5 minute charts. Has anyone found an optimum testing strategy with this short length of data in their experiments ie single walk forward period or multi, what split of in and out sample ratio?
.
