Optimising & % trades not taken

I did an 2 parameter optimise on a weekly system I am working on. The axis marked "Monte Carlo" is a dummy variable - the variable is not used anywhere in the system. The axis marked "trades not taken %" is where a % of buy signals were ignored at random.



With this long term weekly system the more buy signals ignored the lower the CAR (Compound annual return %). I could ignore 10% to 30% of the trades without a huge dropoff in performance.

The "Monte Carlo" axis shows the variation of CAR at different levels of trades not taken. If a big trade is always taken because no trades are ignored then this one trade could give a false impression of the system. By testing with a random buy variable it is possible to approach Monte Carlo simulations using a different randomising approach.

Also I randomised the buy and sell price over the week the signal was given (it's a weekly system) so, even at 100% of signals not ignored there was slight variation in the results.

Hope that all makes sense!

stevo