How to write a CustomFactor for weekly EMA - Python, Quantopian

Question

class EWMAWeekly(CustomFactor):

    inputs = [USEquityPricing.close]
    window_length = (13 + 2 * 13 - 1) * 5  # Initial 13 weeks for sma, then 25 more weeks to improve the accuracy of the current ema.

    def compute(
        self,
        today,
        assets,
        out,
        data,
        ):
        alpha = 2 / (13 + 1)
        weekly_data = data[4::5]  # len = 38, index from 0 - 37
        ema = average(weekly_data[:13])  # Initial SMA
        i = 0
        while i < 25:
            ema = weekly_data[13 + i] * alpha + ema * (1 - alpha)
            i += 1

        out[:] = ema

The CustomFactor above is what I currently have. When I run this through my pipeline, the output is average(weekly_data[:13]), which is the SMA from 25wks ago. The code doesn't raise any errors, and I've tested the while loop, so I know it is running. I believe the problem is with the reassignment of the ema variable inside the while loop. I may be having a moment of stupidity, but I can't seem to find the problem. Any advice is appreciated.

Thanks

Answer 1

It looks like I made a silly mistake. I used ints to calculate the alpha instead of floats. The corrected code is below.

class EWMAWeekly(CustomFactor):

inputs = [USEquityPricing.close]
window_length = (13 + 2 * 13 - 1) * 5  # Initial 13 weeks for sma, then 25 more weeks to improve the accuracy of the current ema.

def compute(
    self,
    today,
    assets,
    out,
    data,
    ):
    alpha = 2.0 / (13.0 + 1.0)
    weekly_data = data[4::5]  # len = 38, index from 0 - 37
    ema = average(weekly_data[:13])  # Initial SMA
    i = 0
    while i < 25:
        ema = weekly_data[13 + i] * alpha + ema * (1 - alpha)
        i += 1

    out[:] = ema