I believe I have finally solved the problem.

class Span_B(CustomFactor):

    #Pre-declare inputs and window_length  
    inputs = inputs = [USEquityPricing.high, USEquityPricing.low]  
    window_length = 104  

    # Compute factor value the span_b line  
    def compute(self, today, assests, out, high, low):  
        for i in range (26,35):  
            span_b_high = 0  
            span_b_low = 0  

            if high[-i-1].all()  > high[-i].all():  
               span_b_high = high[-i-1]  
            else:  
               span_b_high = high[-i]  

            if low[-i-1].all()  > low[-i].all():  
                span_b_low = low [-i-1]  
            else:  
                span_b_low = low [-i]  

            out[:] = (span_b_high + span_b_low)/2  

Hi Zach,

The compute() function can be tricky when you are calculating values over the entire window length like you are here. I think a good place to start is by reading this part of the help documentation. Specifically, this description may help: "On each simulation date, compute will be called with the current date, an array of sids, an output array, and an input array for each expression passed as inputs to the CustomFactor constructor."

Using a for loop in compute() is generally a bad idea as it can make your algo run really slowly, and likely has an equivalent function that can make your desired calculation on the entire matrix of values.

In this case, you might be interested in the numpy.nanmax() and numpy.nanmin() functions. There's an example in the link that I send you that uses these to get the maximum and minimum values over the window of your custom factor. Check it out and let me know if it helps or if you have any other questions.

Jamie,

Numpy.nanmax() and numpy.nanmin will work much better than what I was attempting.

With the window_length parameter is there any way that I could call days from 10 to 20 days ago, or is this limited to only calling the last 20 days. In which case I would need to access the array and select days 10-20 and runing numpy.nanmin/max on that selection. Could this be done by removing the top 10 rows from the array? Or will this interfere with compute?

Thank you in advance,
Zach

Hi Zach,

Take a look at the response by Scott Sanderson in this post. His example shows how to get the mean of the close price, lagged by 10 days, which is very similar to what you're asking!

Let me know if this helps.

Jamie,

That was perfect!

Thank you so much,
Zach

class span_a(CustomFactor):

    #Pre-declare inputs and window_length  
    inputs = inputs = [USEquityPricing.high, USEquityPricing.low]  
    window_length = 52  


    def compute(self, today, assests, out, high, low):  
        from numpy import nanmin, nanmax  

        highest_high = nanmax(high[0: 26], axis = 0)  
        lowest_low = nanmin(high[0: 26], axis = 0)  

        out[:] = (highest_high + lowest_low)/2