Advanced trading software: technical analysis and neural networks

In their article “Zero Lag (Well, Almost)”, John Ehlers and Ric Way have investigated how to remove a selected amount of lag from an exponential moving average and use the filter in an effective trading strategy.

Using Tradecision’s Indicator Builder, one needs to create the ZERO_LAG indicator:

ZERO_LAG indicator:

ins
Length : "Enter Period",20;
GainLimit:"Enter Gain Limit",50;
end_ins

var
alpha:=0;
Gain:=0;
BestGain:=0;
EC:=0;
Error:=0;
LeastError:=0;
EMA:=0;
Value1:=0;
end_var

if HistorySize < 1 then return c;
alpha:=2 / (Length + 1);
EMA:=alpha * Close + (1 - alpha) * EMA\1\;
LeastError:=1000000;
for Value1:= -GainLimit to GainLimit do begin
	Gain:=Value1 / 10;
	EC:=alpha * (EMA + Gain * (Close - EC\1\)) + 
    (1 - alpha) * EC\1\;
	Error:=Close - EC;
	if Abs(Error) < LeastError then
		begin
			LeastError:=Abs(Error);
			BestGain:=Gain;
		end;
end;
EC:=alpha * (EMA + BestGain * (Close - EC\1\)) + 
(1 - alpha) * EC\1\;

return EC;

Using Tradecision’s Strategy Builder, one needs to create the ZERO_LAG strategy:

ZERO_LAG strategy:

Long Rule:

var
Length := 20;
GainLimit:=50;
alpha:=0;
Gain:=0;
BestGain:=0;
EC:=0;
Error:=0;
LeastError:=0;
EMA:=0;
Value1:=0;
Thresh:=1;
end_var

if HistorySize < 1 then return false;
alpha:=2 / (Length + 1);
EMA:=alpha * Close + (1 - alpha) * EMA\1\;
LeastError:=1000000;
for Value1:= -GainLimit to GainLimit do begin
	Gain:=Value1 / 10;
	EC:=alpha * (EMA + Gain * (Close - EC\1\)) + 
    (1 - alpha) * EC\1\;
	Error:=Close - EC;
	if Abs(Error) < LeastError then
		begin
			LeastError:=Abs(Error);
			BestGain:=Gain;
		end;
end;
EC:=alpha * (EMA + BestGain * (Close - EC\1\)) + 
(1 - alpha) * EC\1\;


return CrossAbove(EC,EMA) and 100*LeastError / 
Close > Thresh;


Short Rule:

var
Length := 20;
GainLimit:=50;
alpha:=0;
Gain:=0;
BestGain:=0;
EC:=0;
Error:=0;
LeastError:=0;
EMA:=0;
Value1:=0;
Thresh:=1;
end_var

if HistorySize < 1 then return false;
alpha:=2 / (Length + 1);
EMA:=alpha * Close + (1 - alpha) * EMA\1\;
LeastError:=1000000;
for Value1:= -GainLimit to GainLimit do begin
	Gain:=Value1 / 10;
	EC:=alpha * (EMA + Gain * (Close - EC\1\)) + 
    (1 - alpha) * EC\1\;
	Error:=Close - EC;
	if Abs(Error) < LeastError then
		begin
			LeastError:=Abs(Error);
			BestGain:=Gain;
		end;
end;
EC:=alpha * (EMA + BestGain * (Close - EC\1\)) + 
(1 - alpha) * EC\1\;


return CrossBelow(EC,EMA) and 100*LeastError / 
Close > Thresh;