Objective Function & Optimization

New comment
 
In conducting some walk forward optimizations i'm looking for an efficient yet easy objective function to dictate which settings were the "best". By default MT4 gives you, Profit, Profit Factor, DD$, Expected Payoff, DD%, and Total Trades. By themselves, none of them are mind blowing as an objective function. What about Taking the percentage return and using 100 - DD% to get some sort of ratio. I could use sharpe ratio but I really don't feel like getting into risk free return and programming it manually (probably should stop crying and move to MT5 but hey...). Anyway, any suggestions that are short and sweet?
 

Have a read here.

https://www.mql5.com/en/forum/127371/page2#355175

HTH

V

 
Anyway, any suggestions that are short and sweet?
Definitely sweet.....far from short
 

The challenge, as you've discovered for yourself, is finding a metric whose rank-sorting enables you to effectively downselect the trade parameters of choice which will presumably lead to profits in the future.

Sharpe ratio is flawed, unfortunately, as the deeper analysis of the risk of ruin calcs prove out.

The net of it is that there are no sweet and short metrics. Consider the longevity of financial markets in the history of humankind (thousands of years) and yet even something as simplistic as the sharpe ratio wasn't developed/adopted by the industry until some 30yrs ago.

If you want simple and easy then the quick answers you will get will be useless to your objective. There is no simple and easy. At best it is dang hard but lucrative, at worst this is no better than the fools who spent their lives trying to be alchemist or captain ahab where no matter the determination there was no reward to be gained.

 
Well I guess anything worth doing is worth doing right and I agree with you about the Sharpe ratio. I just wonder if your MAE/MFE is considered the top methodology for analyzing trading systems and settings. Trying to reproduce what you've done, if possible, is months of work and analysis. No doubt its worked for you and your professional endeavors but i'm not sure if that methodology is exactly what I'm looking for (your methods seem very close to the "perfect profit" scenario, which seems to be highly regarded). I guess what i'm trying to do is survey methods used by other professionals, find the one that I feel suits the needs of myself and my trading system, and bring them into practice.
 

IMHO, Accepting any methodology without knowing it's practical applications and understanding its calculation is probably self-sabotaging. Mistaking performance matrix with the Method of a system is the road to madness. Usually, one discover the Method of a profitable system before any performance matrix can help it. We can add Money Management to that list. MAE/MFE is a grading tool, no different from Sharpe Ratio and the count less other methodologies. Together, they give us a better characteristics of a particular trading method. Using them to find the absolute bottom and absolute top of the market will Fail. The more tools within a trader's tool box, the better.

 

One performance matrix used in Blackjack which I don't know would be appropriate for trading would be N0 (N-Zero). It's usually defined as the number to Bets (Trades) needed to over-come 1-One (but I use 3) negative Standard-Deviation. In other-words, it's the number of trades required to be at-worse Break-Even within a 99% Confidence. I wonder if any traders have heard of it? Anyways, I was so happy to see Z-Score being used here. 

 
Bauer_Boy:
Well I guess anything worth doing is worth doing right and I agree with you about the Sharpe ratio. I just wonder if your MAE/MFE is considered the top methodology for analyzing trading systems and settings. Trying to reproduce what you've done, if possible, is months of work and analysis. No doubt its worked for you and your professional endeavors but i'm not sure if that methodology is exactly what I'm looking for (your methods seem very close to the "perfect profit" scenario, which seems to be highly regarded). I guess what i'm trying to do is survey methods used by other professionals, find the one that I feel suits the needs of myself and my trading system, and bring them into practice.

It really depends on your objective. My objective is to use automated market analysis in an effort to forecast the market inasmuch as meterologists forecast tomorrow's weather. As such, in my case it is necessary to have some manner of feedback regarding the accuracy of my strategy's ability to forecast the market.

MAE and MFE provide precisely this kind of feedback. If my strategy says "jump into the market NOW" but the market continues to move against my position then the extent to which my strategy mistimed the market is exactly captured by the MAE. Nothing else gives me that feedback.

Likewise with a determining when to exit a position. A good strategy would predict when the market is about to decidely move against one's position and at such time one would exit from the position. The farther your position is from the MFE the worse your specific strategy is at determining when the market is reversing. Again the best feedback here is the excess MFE.

If you aren't aiming to use automated trading as a means of market forecasting then you might not find much utility in MAE/MFE analysis.
New comment