Hi Mattia,

Thanks for posting and great question. The best way to make everything more realistic is to use a constrained portfolio optimizer to decide weights rather than just placing orders. We have a whole tutorial now that shows you how to do this, or you can also skip to the end and clone the algo if you prefer.

https://www.quantopian.com/tutorials/getting-started

We also have the constraints depend on our risk model, so in addition to constraints that would fix your issue (leverage), there are constraints you can adjust around things like sector and style risk exposure. I recommend keeping the defaults though as we've put a lot of work into determining those AND those are the entry requirements for our upcoming new contest.

Hi Delaney,
This is a backtest in which I contain the Overflow error I mentioned in my previous post.
I really appreciated the hints that you showed me in the tutorial on how to implement a proper risk-management optimisation in my portfolio,
the problem is that when I create my boundaries:

order_optimal_portfolio(  
        objective=objective,  
        constraints=[  
            max_leverage,  
            dollar_neutral,  
            constrain_pos_size,  
            max_turnover,  
            factor_risk_constraints,  
        ])  

I cannot create a proper "objective" in my function because a variable named in this way doesn't work

objective = (for stock in longs:  
        order_target_percent(stock, long_weight)  
    for stock in shorts:  
        order_target_percent(stock, short_weight))  

Is there any other way in which I can fit it in my strategy?
Thanks again for the help
Mattia

Hello Mattia,

The objective is just a vector of weights. Each weight corresponds to your prediction of returns for that stock. If these predictions are correlated with true future returns, that's when you're likely to make money. I go over this here: https://www.youtube.com/watch?v=ArbIM0vhSYQ

This lecture may be very helpful if you want a deeper explanation of all this: https://www.quantopian.com/lectures/risk-constrained-portfolio-optimization

It seems like you just want to make two sets of predictions, one is that a set of longs will go up with uniform probability, and one that a set of shorts will go down with uniform probability. You should be able to achieve this by creating a vector of 1s, 0s, and -1s depending on whether you think a stock is in your longs, shorts, or neither. I'm going to forward this thread to Rob, who may be able to better help with this.

Here is my last result; apparently the strategy works way better with a universe of 1500 stocks instead of 500. Unfortunately I still haven't been able to understand how to apply the "objective" constraint in the for loop.