For code/output blocks: Use ``` (aka backtick or grave accent) in a single line before and after the block. See: http://commonmark.org/help/

Problem trying to port Lowpass filter to backtrader... help!



  • I am trying to port to backtrader an indicator (Lowpass filter), but I am stuck.
    Here you can see the code that works on the usual timeseries (those where
    [0] is the index of the oldest value instead of the current value as is in the
    backtrader's lines):

    #coding=utf-8
    #LOWPASS
    '''
    The lowpass filter is similar to a simple moving average, but with less delay.
    Each elemt of the lowpass (LP) series is a combination of  the last three values
    of the input data (the current one ant the two preceding) and the last two values
    of the same LP series.
    '''
    import matplotlib.pyplot as plt
    import pylab
    import random
    import numpy as np
    import pandas as pd
    import sys
    
    def squarewave():
        wave = []
        for i in range(1,500):
            if (i/100) %2 == 0:
                wave.append(10)
            else:
                wave.append(-10)
        return wave
    
    class LOWPASS(object):
        def series(self, l_data,period):
            a = 2.0 / (1 + period)
            LP = []
            LP.append(l_data[0])
            LP.append(l_data[0])
            swNaN = 1
            for x in range(2,len(l_data)):
                if np.isnan(l_data[x-2]):
                    LP.append(np.nan)
                else:
                    #print x
                    if swNaN == 1:         #initializes LP vith the first element (not NaN) of data series.
                        swNaN = 0
                        LP[x-1] = LP[x-2] = l_data[x]
                    LP.append((a - 0.25 * a * a) * l_data[x] + 0.5 * a * a * l_data[x-1] \
                    - (a - 0.75 * a * a) * l_data[x-2] \
                    + 2 * (1. - a) * LP[x-1] \
                    - (1. - a) * (1. - a) * LP[x-2])
            return LP
    
    if __name__ == '__main__':
        mylist = squarewave()
        myLOWPASS = LOWPASS()
        #mylist=[10,20,30,40,50,60,70,80,90,100,101,102,103,104,105,106,107,108,109,110]
        #mylist= range(1,500,10) + range(501,1500,1)
        #mylist = []
     
        #for i in range(0,1000):
        #    mylist.append(random.randint(1,101))
        #    mylist.append(100*np.sin(i))
       
        print mylist
        lowpassed = myLOWPASS.series(mylist,25)
        print "lowpassed: ", lowpassed
        df = pd.DataFrame()
        df['mylist']= mylist
        df['lowpass']=lowpassed
        print df
        df.plot.line()
        plt.show()
    

    The problem I have when converting this code to backtrader is with the initialization of the LP line:

      LP.append(l_data[0])
      LP.append(l_data[0])
    

    In backtrader I do not know how to set the first two elements of the LP line to the oldest value of the data
    line: obviously my understanding of backtrader is too limited... any help will be appreciated.


  • administrators

    backtrader uses index 0 as the pivot point for the current data, i.e.: the index to which your indicator writes values (indicators which have been calculated prior to yours, will have placed data also at index 0)

    Index -1 is the one where the last value was placed.

    See the reference: Docs - Platform Concepts

    To get the farthest point (first data in the series) from the current one:

    • If len(data) == 1 the last point is at 0
    • If len(data) == 2 the last point is at -1

    Using that:

    first_data_index = -len(data) + 1
    


  • Thanks. Nice to see you again on this blog, I was starting to think that you intended to abandon the project!