Robustness is really important for trading systems as it determines whether they can stand the test of time. I understand that many systems developers out there do parameter stepping to optimize. Through optimization, they find a set of values that yields the best return with the lowest drawdown. I am not going to criticize but be warned that it used excessively, its called overfitting. Your system will have a higher probability of breaking down in the long run.
Another way to determine the robustness of your system through parameter stepping is to test out certain values and see if there are spikes in your chosen evaluation method. For me, my evualuation method is based on MAR and annualized return.
Whenever I test a system, I arbitrarily choose a set of parameters (off my head or fib numbers). I program the entire system and then I do a parameter stepping to see if the values I chose are robust or not. I define robustness of a trading system by the performance of surrounding parameters. If performance is very similar and don’t deviate from the original parameter performance, then I am more confident that adverse changes in market conditions will not affect my system greatly.
Below are graphs of my parameter stepping of the long term trend filter I use. The stepping starts from 150 days lookback all the way up to 350 days in increments of 5. Tested on a diversified basket of commodities futures from 1989 to 1995. Round lot commision = $75.