I generally optimise as per your 1) Test indicator parameters first, then stop loss, and then target profit?
When I optimise 2 or more parameter at the same time, it can take so long I get frustrated! 🙂
If optimising 2 or more parameters together, one trick is to do a 1st run with large increments … 10, 20, 30 etc and then do a 2nd run centred around the figure from the 1st tun. So if you get 20 on 1st run then you may want to optimise min 10 max 30 at increments of 2,4,6 etc. At low incremental values you may be curve fitting?
You need to peruse the optimised profit figures and make sure that there isn’t a sharp rise over one or two parameter variables. If there is a sharp rise then you probably have curve fitted to some odd scenario (e.g non farm payroll day or UK Brexit Vote day etc etc).
When looking at the optimum figure for variables, you want a gradual increase & decrease in profit value either side of the optimum figure. So you may not always use the highest profit figure for variable (due to curve fitting). Hope I’ve explained that to make sense?
I generally optimise over a period covering 100,000 bars and if the instrument price curve (not equity curve) covering the 100000 bars shows plenty of long ups & downs, short ups & downs, flat-ish periods followed by sharp corrections etc. AND also the optimising produces > 500 trades or better still > 1000 + trades.
If the Bot runs 24 hours then I enter a spread figure in the optimiser which is an average of day spread and night spread.
Optimiaing over lots of ‘Ups & Downs & Flats & high number of trades’ with results giving acceptable profit at acceptable risk & drawdown means the Bot has passed my checks.
I then launch the Bot in Demo mode as a real time walk forward test.
AND STILL THEY DON’T WORK! ha (joke!)
GraHal
.
