A simple solution to my problem would be if I could change the day counter in such a way that it counts only the business days since an earnings release. Can anyone suggest a way to do that:. My approach below produces an error:

TypeError: Iterator operand 1 dtype could not be cast from dtype('<M8[ns]') to dtype('<M8[D]') according to the rule 'safe'

class DaysSinceRelease3(CustomFactor):  
    # Only getting the previous quarter's estimize surprise  
    inputs = [fe.Actuals.slice('EPS', 'qf', 0).asof_date.latest]  
    window_length = 1

    def compute(self, today, assets, out, input_1):  
        days =  today.tz_convert('US/Eastern') -  pd.to_datetime(input_1[-1, :]).tz_localize('US/Eastern')  
        #days =  today.tz_convert('US/Eastern') -  pd.to_datetime(input_1[-1, :]).tz_localize('US/Eastern')- timedelta(days=1)  
        #start = dt.date( today)  
        #end = dt.date( pd.to_datetime(input_1[-1, :]) )  
        start =  today  
        end = pd.to_datetime(input_1[-1, :])

        days = np.busday_count(start ,end)  
        out[:] = days.astype('timedelta64[D]')  
        #out[:] = days  

        #pipeline_output.index.set_levels(pipeline_output.index.levels[0]- timedelta(days=1),level=0, inplace=True)

Check out this other post https://www.quantopian.com/posts/count-the-number-of-business-days-since-earnings. There is a built in factor for business days since event. No need for a custom factor. The factor is BusinessDaysSincePreviousEvent (https://www.quantopian.com/docs/api-reference/pipeline-api-reference#quantopian.pipeline.factors.BusinessDaysSincePreviousEvent).

That should do what you want.

Do you think it would be possible to make the asofDate match the index date exactly? Currently, there is always a 1 day difference between the as of date and the date in my index column (maybe to avoid look ahead bias??)