Hello, Apologies as I've asked a similar question before;
It is suggested that if a system is more robust (less curve fitted), its' performance shouldn't be affected too much by the addition of noise. I have coded Howard's "monte-carlo" code into the code of a trading system and optimised the noise. I've ranked the results by the increasing addition of noise and the following happens; The net return % decreases as noise is added The RAR increases as noise is added (basically less trades are taken, but the quality of them increases) as noise is added My custom metric (a factor of Kelly, RAR, CAR/MDD an Net% profit) increases with increased noise. My question - what performance stat should I be comparing to the noise, and is it realistic that my system's performance should IMPROVE with the addition of noise, by these measures? All contributions welcome. Cheers, Rich