[Stefan Scherbarth]

Great spreadsheet, but some remarks about the concept:

The Portfolio Builder optimizes a portfolio with hindsight on mainly backtested system curves also created with hindsight. Therefore like any heavy system optimization the outcome is most likely not indicative for the future performance of the portfolio. Sometimes a heavily optimized solution turns out to be worse than a not optimized one.

To at least get an idea what the (hopefully positive) effect of the portfolio optimization is, I suggest to divide the time considered into an optimization period and a test period ( According to some standard procedure in system development).
Even then the result may be overfittet as long as the underlying system time series result from a back-test optimization, so the most strict approach would allow only real forward results without strategy change to be used to calculate the portfolio result. Unfortunately the periods are so far quite small for a good statistics, however I suggest to also indicate and ideally graphically mark the date where the respective portfolio shows real forward tested result.

Regarding draw down the maximum draw down is a quite bad quality indicator as it heavily depends on the time span selected and the events included. The monte carlo simulation of the equity curve is a well known and much better approach to fully exploit the information contained in a time series. Therefore it would be great if the calculation of the maximum portfolio monte carlo drawdown at a selectable percentil of the monte carlo time series would be available.

[Author]

Posts