Optimizing equity curve trading

Posted By: Brax

Optimizing equity curve trading - 08/21/17 17:14

Hi,

I am trying to optimize the lowpass filter in equity curve trading for a basket of assets. It seems the optimization is not being performed, surely there is some kind of issue with Balance or ProfitClosed behaviour. This is my code:

Code:

static var opt_Filtered;
static var opt_Signal;
static var opt_Threshold;
static var opt_Stop_Cntr;
static int opt_EquityAvg;

var equityCurveSizing()
{
  if(Train) { return 1; } // no phantom trades in training mode
  vars EquityCurve = series(BalanceLong+BalanceShort);
//  vars EquityLP = series(LowPass(EquityCurve,10));
  vars EquityLP = EquityCurve;
  if(EquityLP[0] < LowPass(EquityLP,opt_EquityAvg))// && falling(EquityLP))
    return -1; // drawdown -> phantom trading
  else
    return 1; // profitable -> normal trading
}
	
function tradeCounterTrend()
{
	vars Price = series(priceClose());
	vars Filtered = series(BandPass(Price,opt_Filtered,0.5));
	vars Signal = series(Fisher(Filtered,opt_Signal));
	var Threshold = opt_Threshold;
	
	Stop = opt_Stop_Cntr * ATR(10);
	Trail = 4*ATR(10);
	
	if(crossUnder(Signal,-Threshold))
		enterLong(); 
	else if(crossOver(Signal,Threshold)) 
		enterShort();
}

function run()
{
	set(LOGFILE+PARAMETERS);
        NumCores = -2;
	BarPeriod = 24*60;
	LookBack = 500;
	StartDate = 2005;
	EndDate = 2015;
	Capital = 10000;
	
	if(ReTrain) {
		UpdateDays = -1;
		SelectWFO = -1;	
		reset(FACTORS);	
	}

	while(asset(loop("EUR/USD","USD/JPY","GBP/USD","USD/CHF"))){
        
        // Parameters	
	opt_Filtered = optimize(30,20,40,1);
	opt_Signal = optimize(10,5,20,1);
	opt_Threshold = optimize(1,0.5,1.5,0.1);
	opt_Stop_Cntr = optimize(4,2,10,1); 
        opt_EquityAvg = optimize(100,5,200,5);

        Lots = 5;
        Lots = equityCurveSizing() * Lots;
	tradeCounterTrend();	
	}
			
	PlotWidth = 1024;
	PlotHeight1 = 400;
}

�The behaviour of equity curve trading is for the global equity produced from all the assets or it�s applied on every asset equity?

If globally, �Could i store the different asset equities and apply it separately with a function that receives that equity?.

Thanks.

Posted By: Brax

Re: Optimizing equity curve trading - 09/08/17 20:00

Well, after some research i can answer myself some questions:

First, equity curve trading as explained in the manual is only for global equity produced by all trades from all assets. If you want to individualize it, you must do some workarounds.

Second, i haven�t realized how to optimize the parameter for the average, but it doesn�t matter.

Finally, i have found a mechanism i think is better than doing it with averages. It consists of applying the kelly factor for every asset on every trade close, this way i don�t need any parameters at all and it serves the purpose of avoiding broken systems to be traded.

The only problem is that most times it hinders performance but seems to be a good contingency plan.

More about this can be found here:
http://www.financemagnates.com/forex/bloggers/equity-curve-trading-part-2-rolling-expectancy-ratio/

In my case, as i don�t want to add more parameters i�ve used all data to implement it. Hope it is useful for everybody.

Posted By: MatPed

Re: Optimizing equity curve trading - 09/09/17 16:59

Well, I do not think that what you stated is correct. I do check equity curve for each Algo/Asset Combination.
Just call the checkEquity() function in each Algo script before opening the trade and the job is done.
In order to size the trade I use the Margin not LOTS in order to evoid conflicts between sizing the trade and equity line check.

My 2 cents. Ciao

Posted By: Brax

Re: Optimizing equity curve trading - 09/10/17 20:24

Hi MatPed.

Surely you are right, i�ve always avoided to place optimize statements directly in algo scripts, as i tend to do it at a global level. Probably that�s the reason why i couldn�t optimize the average...

I�ve used lots instead of Margin just for simplicity, but it�s good to know that could be another reason why i couldn�t optimize equity curve trading properly.

Anyway, i see a problem with doing it with averages, this is because we add more parameters to the equation, so we unintentionally increase curve fitting.

You can check the first part of the article and see what i mean, i think applying expectancy or kelly factor over the whole data is a good way of avoiding disasters without adding parameters nor bias. The rolling versions work in a similar way like doing it with averages.

There is people who even just stop trading when they reach MDD95 from Montecarlo.

Thanks.