Is it possible to dynamically change a builting function parameter?

[demarcog]

Normally function definitions such as:

" vstochK = stochK(Length,3,3)"

are defined within the bInit (execute once) portion of a EFS routine and executed once per EFS invocation.

Once defined the values generated by the stochK() function are retrieved via the vstochK.getValue(-1) call.

Is it possible to alter the Length value within the EFS outside of the bInit section to pass different values to the stochK function. For example for one bar passing a 14 and another bar passing 21, etc?

When I place the function definition outside the bInit loop the EFS backtest takes a very long time and frequently runs out of memory or crashes eSignal.

Has anyone tried doing this?

thanks very much,

Glen

[ACM]

Re: Is it possible to dynamically change a builtin function parameter?

Glen
It is possible however for each new set of paramaters the efs engine will create a new instance of the series object that will remain in memory for the life of the formula and will eventually fill up the available memory
Alex

Originally posted by demarcog
Normally function definitions such as:

" vstochK = stochK(Length,3,3)"

are defined within the bInit (execute once) portion of a EFS routine and executed once per EFS invocation.

Once defined the values generated by the stochK() function are retrieved via the vstochK.getValue(-1) call.

Is it possible to alter the Length value within the EFS outside of the bInit section to pass different values to the stochK function. For example for one bar passing a 14 and another bar passing 21, etc?

When I place the function definition outside the bInit loop the EFS backtest takes a very long time and frequently runs out of memory or crashes eSignal.

Has anyone tried doing this?

thanks very much,

Glen

[demarcog]

Alex,

Thanks for the response it sounds like that is exactly what I am seeing happening.

Would another alternative be to include the code for stochK() inline then just modify that one variable rather then generate the entire series each time?

Is the code for the buildin functions generally available as I don't recall seeing the EFS code for the stochastics calculation?

If you know of a file share that has a close approximation I would appreciate you letting me know. I thought I recalled you posting the William Blau's version in an EFS somewhere. I ran some back tests awhile ago comparing the crossovers between Lane and Blau's and was surprised that there was not a significant difference.


Here is Lane's formula:

Stochastics Formula
A momentum or price velocity indicator that was developed by George C. Lane. It represents a fixed period-to-period moving calculation that is very susceptible to instability and false signals since it is possible for the indicator to fluctuate wildly by simply removing data from the oldest period.


K = (C -L)
_____ * 100
( H - L )

K = Lane's Stochastics
C = the latest closing price of the stock or contract
L = n-period low price of the stock or contract
H = n-period high price of the stock or contract
n = any number (Lane recommends a range of 5 to 21)
SK or %K = three-period simple moving average of K
SD or %D = three-period simple moving average of SK

Source: The Encyclopedia of Technical Market Indicators, Colby and Meyers


Thanks again Alex,

Glen

[ACM]

Glen

Would another alternative be to include the code for stochK() inline...

Not sure what you mean by "inline".
If you mean retrieving the value of the stochK() directly [ie stochK(lengthK, smoothK, smoothD, 0)] then this will also create a new instance of the series object each time you pass a new set of parameters.
If instead you mean using a custom stochastic function that you call as required then that could be an alternative.
Enclosed below is an old Stochastic%K formula I had that you may want to use
Alex

function main(lengthK,smoothK) {
    if(lengthK == null) lengthK = 14;
    if(smoothK == null) smoothK = 1;
    var nCount,Close,LL,HH,percentK;
    var nSum = 0;
    for(var i = 0; i < smoothK; i++) {
        nCount = i;
        Close = getValue("Close",-nCount);
        LL = StochasticLL(lengthK,nCount);
        HH = StochasticHH(lengthK,nCount);
        if(Close == null || LL == null || HH == null) return;
        if((HH-LL)!=0) percentK = ((Close-LL)/(HH-LL))*100;
        nSum += percentK;
    }
    nSum /= smoothK;
    return nSum;
}
function StochasticLL(inputLength,inputCount) {
    var LL = 0;
    var Low = getValue("Low",-inputCount,-inputLength);
    if(Low == null) return;
    for(var i = 0; i < inputLength; i++) {
        if(i == 0) {
            LL = Low[i];
        } else {
            LL = Math.min(LL,Low[i]);
        }
    }
    return LL;
}
function StochasticHH(inputLength,inputCount) {
    var HH = 0;
    var High = getValue("High",-inputCount,-inputLength);
    if(High == null) return;   
    for(var i = 0; i < inputLength; i++) {
        if(i == 0) {
            HH = High[i];
        } else {
            HH = Math.max(HH,High[i]);
        }
    }
    return HH;
} 

Originally posted by demarcog
Alex,

Thanks for the response it sounds like that is exactly what I am seeing happening.

Would another alternative be to include the code for stochK() inline then just modify that one variable rather then generate the entire series each time?

Is the code for the buildin functions generally available as I don't recall seeing the EFS code for the stochastics calculation?

If you know of a file share that has a close approximation I would appreciate you letting me know. I thought I recalled you posting the William Blau's version in an EFS somewhere. I ran some back tests awhile ago comparing the crossovers between Lane and Blau's and was surprised that there was not a significant difference.


Here is Lane's formula:

Stochastics Formula
A momentum or price velocity indicator that was developed by George C. Lane. It represents a fixed period-to-period moving calculation that is very susceptible to instability and false signals since it is possible for the indicator to fluctuate wildly by simply removing data from the oldest period.


K = (C -L)
_____ * 100
( H - L )

K = Lane's Stochastics
C = the latest closing price of the stock or contract
L = n-period low price of the stock or contract
H = n-period high price of the stock or contract
n = any number (Lane recommends a range of 5 to 21)
SK or %K = three-period simple moving average of K
SD or %D = three-period simple moving average of SK

Source: The Encyclopedia of Technical Market Indicators, Colby and Meyers


Thanks again Alex,

Glen