Hi Angel,

Typically, for this type of help, it's much easier if you share your code so that we the community can see what you're doing and potentially suggest a fix or improvement. If you don't feel comfortable sharing your strategy, perhaps you could create a copy that highlights the problem without revealing important parts of your algorithm?

My suggestion for a 3-day mean Returns factors would be to use Pipeline. Pipeline allows you to create factors which you can then use to select and weight securities. Check out the Pipeline Tutorial. In particular, I think lessons 3 and 4 might be helpful.

Hello Jamie,
Many thanks for your suggestion, much appreciated. I will definitely share this weekend the code. It's literally a few lines of code.
I thought to ask you because perhaps other people had already coded something similar.
If in the meantime you could also shed some light, all I'd really need is the code to calculate a simple moving average of the last X daily returns.
Even the simplest of the code would be of great help.
Thanks,
Angel

Hi,
this is the code that, as mentioned, I copied from the post above.
I think the code isn't right, but I can't figure out why.
Thanks!

import numpy as np  
import datetime  
import pandas as pd

def initialize(context):  
    context.security = sid(8554) #spy  
def handle_data(context, data):  
    context.exchange_time = pd.Timestamp(get_datetime()).tz_convert('US/Eastern')  
    hist = data.history(context.security, "close", 3, "1d")  

    # Getting the daily returns  
    simple_returns = hist.pct_change()  
    log_returns = np.log(hist / hist.shift(1))  
    # Getting the average return  
    avg_simple_return = simple_returns.mean()  
    avg_log_return = log_returns.mean()  
    if context.exchange_time.hour == 15 and context.exchange_time.minute == 59:  
        print "3 most recent close prices:"  
        print str(round(hist[-3],3)) + ", " + str(round(hist[-2],3)) + ", " + str(round(hist[-1],3))  
        #print str(round(avg_simple_return[-1],6)*100) + "%"  

    # Standard deviation of returns  
    stdev = simple_returns.std()  

Angel, your code looks close, with one big problem. And there are a handful of things that can be better.

• If you want the returns of 3 days, you actually need 4 days worth of close prices. The returns calculation depends on 2 days of data, and if you try to compute a return with only one price, you get a NaN.
  
• Don't run things in handle_data() unless you need it every minute. It slows your backtest down. Use schedule_function() instead.
  
• You have standard deviation in your code, but you don't use it. Delete the useless stuff.

I'm attaching a backtest for you to look like. (It's much easier to share backtests than copy-and-paste code). It has some better coding practices and it logs and records the data you're looking for. You still have to add the ordering logic. You can choose the time of day for the orders by editing the schedule function.

Dan, thank you so much, this is very useful and I'll be more careful in the future, sharing/attaching directly my backtest file.
Angel