Issue with reults of rma moving average

Describe the bug ta.rma() should behave exactly like ta.ema() - with the only difference being the weighting factor. Implementation of ta.rma() in pandas_ta calculates the starting value differently than the implementation of ta.ema() specifically, the first value in the ta.rma() result should be a SMA() average of the period. If the first value is different, the whole series is different…

To Reproduce

df = pd.DataFrame({'close': [1.0,0,-1,0,2,-3,4,5,-8,7,8,9,10,-5,10,0]})
df["rma"]=df.ta.rma(close=df["close"], length=4)
df["ema"]=df.ta.ema(close=df["close"], length=4)

the first value that is not a NaN should be the same for both rma() and ema() - but it is not…

[mihakralj]

Sounds good. 😊 I did find this discrepancy by getting different data from ATR (average true range) compared to the official example by J. Wilder (the author of both ATR and RMA). Let’s keep the library as it is then, if all other rma-derived indicators are satisfactory for library users. I would not mess with adding another parametric option to make the indicator following the author’s original formula, let’s keep the current simplified version.

[mihakralj]

Yup, can confirm that you found the data bug in a 42-years old book! 😃 Below is the calc of TR and ATR for the same raw data from the book using my C# lib: