Yes. I agree, they are not the same thing. _BUT_, you should have a "penalizing function" everytime you look at the same data - otherwise, it is _YOU_ who is doing the optimization by designing ever more accurate systems to better "fit" the patterns you see. nitro
Agree w/ the above, but I think it highlights the discretionary trader's viewpoint more than anything else. The discretionary trader doesn't confine his flexibility with rigid rules precisely because he trusts himself more than he trusts his rules. That said, anyone who wants to hand their decisions over to a system- trade mechanically- is taking the completely opposite tack: not trusting themselves for whatever reason, and putting complete faith in their rules instead. Direct opposite. This life and limb rigid rule dependence creates a dilemma: -backtesting the rules is fraught with peril. -NOT backtesting the rules is fraught with peril. No easy answer except to know what you're doing and be good at what you do... The discretionary trader has to overcome his emotions and psychological weaknesses; there is no way forward otherwise. The systems trader intentionally avoids the discretionary trader's human issues- but in doing so he forfeits the benefits of human flexibility and human intuition, thus creating a whole new set of automization issues and 'dumb computer' issues. This is only fair of course. No free lunch
I propose this question with sincere respect... Mechanical System trading critics: I have constructed a system (in Tradestation) and backtesting shows $20K return last year on one ES contract. 900 trades = $4,320 in commissions via IB. Trade volume is consistent indicating no short/long bias. It looks to me like I should be net $15,680 per contract. What am I missing here? (Thank you all for your thoughtful input into this thread)
1) How much optimization and how many times did you try a system on the data until you settled on this one. 2) Post all the statistics that TS gives you, e.g, drawdown, number of trades, max win, max loss, etc etc. nitro
Two things come to my mind. How many parameters are you using? A sure fire way to tell if you have overoptimized the system is if you have an excess use of parameters (>3 of 4 ). Some hedge funds actually use systems with 1 or 0 parameters. Now those are truly robust systems. Second and more importantly have you tested that very same system (no parameter changes) on different markets / timeframes? There should be minimal degradation between markets/most timeframes (ie naturally some systems breakdown when you are zoomed in too much and the noise level becomes too high). Personally I would not trade any mechanical system unless it works on all markets and many different timeframes. Hope this helps. PM me for more details if you wish.
1.) This is about the 3rd incarnation of this system applied to TRADESTATION. It's my first time backtesting via software. I used to just keep detailed logs in EXCEL... 2.) Stats to come... 3.) Oh, as you will see the actually gross is just shy of $30K, not $20K. Didn't want to sound too extreme.