In my on going quest to find the building blocks of a better, more adaptive Parabolic Search and Rescue Stop and Reversal I was coached by David Varadi of CSS Analytics and Jeff Pietsch, aka Market Rewind, to run a Walk Forward Test on several assets from different asset classes using several performance measures. I chose Omega, Sortino Ratio, Skewness, Win/Loss Ratio and total average of each year’s average daily return.

So I picked the 10 Year Treasury future:

Gold Future:

Corn Future:

Euro, Swiss Frank & Australian Dollar Futures:

, ____ ,____

and E-Mini S&P 500 Future:

The Good, the Bad and the Ugly Consistencies:

  • Omega is just flat out dead at 0 or very close to it.
  • Sortino is also ugly: consistently a steady close to -1 every time.
  • Total (17 year) Average of Annual Skewness surprisingly positive on all parameters in every test and the bigger the parameter the less positive the skewness but the worse the long term performance (simple cumulative percentage returns).
  • Total average of each year’s Win/Loss Ratio also increases as the parameter increases.
  • Total average of each year’s average return: All instruments had the highest average return in the lowest parameter besides S & P 500.
  • The Simple (non-compounded) Cumulative Performance has been been consistently better at the highest 2 parameter settings 0.09 and 0.10.

The ugliest and most inconsistent asset in the bunch was ES_F: S & P 500 future. The first few parameters actually produced a negative simple (not compounded) cumulative performance unlike all other instruments. It also took a relatively big hit in ’08 compared to all other instruments: the smaller half of the parameter settings from 0.01 to 0.10 (with the exception of the default 0.02) were actually in the red. Unlike all the rest, its average of each year average return increased as the parameter increased and the lowest parameter, 0.01, had the lowest performance.

Conclusions:

On one hand, it’s pretty clear that 0.09 and 0.1 should be the only parameters I should be basing the further development stages on. The higher the parameter setting the bigger the slice of the trend range pie PSAR gets… (up to a certain point, of course. I only tested it up to 0.1).

On the other hand, even if one just looks at the top 2 performing parameters, they’re still disasters through the eyes of Sortino and Omega.

On a side note, it was (and still is) a great lesson in programming in VBA, learning Excel’s intricacies and even though I’m just scratching the very edge of the surface of what it takes to build profitable strategies, I still very much enjoy it.