true_range() function are not propagating drift parameter in the inputs - NaN values.

Which version are you running? The lastest version is on Github. Pip is for major releases.

import pandas_ta as ta
print(ta.version)
0.2.45b0

Describe the bug The true_range calculation is valid when its inputs ([high_low_range, high - prev_close, prev_close - low]) are valid. This assumption is not occurring in the beginning of the data series, due shift caused by drift parameter.

The true_range() function has the argument drift which is a shift in the data series. This causes introduction of NaN (depends on numbers of shifts) in the beginning of the data serie close_prev. So, every calculation that relies on first series are using NaN as input. The bug is that true_range() assumes NaN as a valid input, so the results are compromised to a partial calculation, which I think it is incorrect. This error propagates to subsequent calculations, for instance average.

To Reproduce Inputs: testDataframe.zip

if __name__ == '__main__':

    # read input data with expected results
    # columns are: close, close_prev, high, low, true_range, true_range_expected_result, ATR_SMA_20_expected_result
    filenamePath = r'D:\scripts\trade\ta_analysis_pr\testDataframe.csv'
    df = pd.read_csv(filenamePath, sep=';')

    # calculate true range for all rows
    df.ta.true_range(append=True)
    # calculate SMA_20 based on true_range
    atr = ta.sma(df.TRUERANGE_1, length=20)
    # join SMA result
    df = df.join(atr)

    print(df[['close_prev', 'true_range_expected_result', 'TRUERANGE_1']])
    # compare columns true_range_expected_result vs. TRUERANGE_1
    # check df.TRUERANGE_1[0]=0.32, true_range() is returning a partial value, because df.close_prev[0]=NaN, i.e., the function doesn´t have all valid inputs (close, close_prev, high, low) to provide a valid value.
    print(df[['TRUERANGE_1', 'ATR_SMA_20_expected_result', 'SMA_20']])
    # compare columns ATR_SMA_20_expected_result vs. SMA_20
    # check df.SMA_20[19]=0.9495, the result is partial because true_range assumed close_prev as NaN in the first cycle. Due previous error, the sma calculated incorrect.

    df.to_csv(r'D:\scripts\trade\ta_analysis_pr\testDataframe_result.csv', index=True)

Expected behavior The attachment shows an example of calculation of arithmetic average calculation based on true_range results. testDataframe_result.xlsx

The changes in close_prev due shift (introducing NaN) should propagate to inputs of calculation of: true_range = true_range.abs().max(axis=1) I did a proposed solution in: https://github.com/ffhirata/pandas-ta/tree/ffhirata-patch-1

Screenshots If applicable, add screenshots to help explain your problem.

Additional context Add any other context about the problem here.

Thanks for using Pandas TA!

[ffhirata]

Happy to help!

[twopirllc]

Hello @ffhirata,

This should do the trick. It would be great if you test it and make a PR so you can be added as a contributor.

from numpy import nan as npNaN # include

def true_range(...):
    # ...
    true_range = true_range.abs().max(axis=1)
    true_range.iloc[:drift] = npNaN # << fix
    # ...

Thanks, KJ