PowerBASIC Indicators

[Nate]

I have taken some time in recent months to learn PowerBASIC, and specifically for creating DLL’s that expand upon the industry leading offering of Neuroshell built-in and custom indicators. One of the areas of interest for me is normality testing (ie. time series data distribution analysis).

I am by no-means a statistician, but John Ehlers article, “Probability—Probably A Good Thing To Know” (Oct, 2018 – TASC) discusses the importance of normally distributed mean reverting indicators and trading strategies. Ward Systems took the time to create a Neuroshell indicator that exports the binned dataset to an Excel csv file, which can be visually studied for normality characteristics (Ward TASC Indicator).

Rather than boring the group with another “Scholarly” article, I instead have included the PB code below. This code implements a simple rolling/moving Jarque-Bera (JB) statistic calculation. As opposed to visually interpreting the output, this indicator can be used as a rule-based interpretation of the time series. Simple internet browser searches will allow you to arrive at explanations for this JB normality test.

If you are interested in expanding your horizons as it pertains to Normality Testing, the following article breaks down the basics:

Testing For Normality

PowerBASIC Code:

SUB JB_Test ALIAS "Jarque-Bera" (   BYVAL PRC_input AS    DOUBLE PTR  , _
                                                BYVAL Per           AS      LONG            , _
                                                BYVAL signal    AS    DOUBLE PTR  , _
                                                BYVAL cnt       AS    LONG        ) EXPORT
    REGISTER i&, j&
    DIM s_ave#, s_sdev#, s_var#, s_skew#, s_kurt#, missing#
    DIM s_sum#,  s_diff#, s_p#, s_adev#, s_sumsq#, s_sumsk#, s_sumku#, s_JB#
    DIM cf_1#, cf_2#, cf_3#     ' Cheater Functions
    missing=3.4e38

   FOR i=0 TO cnt-1  ' go through the entire chart
        IF i < Per-1 THEN   ' there are not yet enough of the time series for the first average
            s_var = missing
        ELSE
            'Step 1 - Calculate Windowed (Period) Mean
            s_sum=0
            FOR j=i-(Per-1) TO i
                s_sum = s_sum + @PRC_input[j]
            NEXT j
                s_ave = s_sum / Per

            'Step 2 - Calculate Windowed Variance & StDev
            j=0
            s_sumsq = 0
            FOR j=i-(Per-1) TO i
                cf_1 = (@PRC_input[j] - s_ave)
                s_sumsq = s_sumsq + (cf_1 * cf_1)
            NEXT j
                s_var = s_sumsq / (Per - 1)
                s_sdev = SQR(s_var)

            'Step 3 - Skewness
            j=0
            s_sumsk = 0
            FOR j=i-(Per-1) TO i
                cf_2 = ((@PRC_input[j] - s_ave)/s_sdev)
                s_sumsk = s_sumsk + (cf_2 * cf_2 * cf_2)
            NEXT j
                s_skew = s_sumsk * (Per / ((Per-1)*(Per-2)))

            'Step 4 - Kurtosis
            j=0
            s_sumku = 0
            FOR j=i-(Per-1) TO i
                cf_3 = ((@PRC_input[j] - s_ave)/s_sdev)
                s_sumku = s_sumku + (cf_3 * cf_3 * cf_3 * cf_3)
            NEXT j
                s_kurt = s_sumku * (Per * (Per + 1) / ((Per - 1) * (Per - 2) * (Per - 3))) - (3 * (Per - 1) ^ 2 / ((Per - 2) * (Per - 3)))

            'Step 5 - Jarque-Bera Test Calcs
                s_JB = Per * ((s_skew * s_skew)/6 + (s_kurt * s_kurt)/24)
            @signal = s_JB
        END IF
   NEXT i
END SUB          

Nate

[Nate]

AS a follow-up to the prior post, attached is a Zip file containing the following items:
– Example NSDT Cht File
– Compiled DLL

***
I tried to post these files, however had to edit with this note given the forum appears to restrict posting certain file types. I have submitted a request to customer support and will post as soon as I get a response
***

Nate

[Ward Systems Group Support]

The forum now has the ability to attach .cht and .zip files.  However, .DLL files are not allowed to be posted.

 

Attachments:

You must be logged in to access attached files.

[Nate]

As promised, please find attached the ZIP file containing the aforementioned indicators.

Regards,

Nate

Attachments:

You must be logged in to access attached files.

[Nate]

Arnaud Legoux Moving Average (ALMA)

I have spent some time over the last couple of days analyzing some smoothing indicators (Moving Averages), and stumbled across some discussion regarding the ALMA indicator.  This is one of the newest additions (2009) to the the moving average family and shows some really good characteristics.

I noticed that there was an ALMA filter already within my template file, however it’s performance is far different than the one that I coded.

In an effort to visualize the performance of the indicator, attached you will find a chart export from Neuroshell.

The charting is broken down into two views:

• Performance for artificial impulse (20 per, .5 amplitude)
• Performance for EOD stock ticker

Also attached is a ZIP file containing the compiled indicator DLL, BAS file, CHT file and TPL files.

As a final note, I also attached the Original document describing the development of the ALMA indicator.

Regards,

Nate

Attachments:

You must be logged in to access attached files.

[tradercleve]

Very nice…ALMA performance jpg…try also Cybenetic Analysis Netlead with Price, Close and Open compare those three to the ALMA indicator. I use the three Netlead indicators for eyeballing potential entry/exits…but I will have to look into the ALMA indicator some more.

I will try to post a chart here with that info within this next week.

I am looking for FOREX trading entry strategies to hold for a few hours to a day or two.

Trader Cleve