Here is the code and the backtest results: From 2002-01-03 to 2014-12-28 6417.8% returns and 49% max drawndown.

Some tools and changes to consider. Try to avoid negative cash.

Hey, Python/SQL newbie here, so this might be a silly question regarding your code, but when I'm looking at the do_screening method

.filter(fundamentals.cash_flow_statement.operating_cash_flow > fundamentals.income_statement.net_income)
        .filter((fundamentals.valuation.enterprise_value / fundamentals.cash_flow_statement.free_cash_flow) < 15)  
        .limit(context.max_num_stocks)  
    )  

the .limit query modifier appears to be for the purpose of limiting the amount of stocks in your portfolio to the maximum defined in the initalize method.

However, I don't understand what the .limit behavior is in the absence of an order_by modifier. IE, I assume there is an inherent ordering off of which the max_num_stocks cuts off of, but it's unclear to me whether that's based on the data order in Morningstar's database, or an order shift based on how the queries are run. Can you tell me how this ends up working int he context of the algorithm or where to look in the documentation?

Thanks.

Hi Don,

you're right, .order_by is essential if you use the .limit clause. The code and backtest above are quite old. I've completely redesigned my implementation of the Millennial Money Algorithm und maybe I'll publish is in the next time.

Please be aware that this strategy is almost 2000% laverage.

I just took a crack at making a simple hedged portfolio inspired by this algo. (I used the algo shared above by @garyha as the foundation). I simply short an equivalent amount of SPY (e.g. dollar-neutralize the portfolio) at each quarterly rebalance period. It does quite a good job of bringing the beta down to nearly 0 over the whole backtest as well as reducing the Max Drawdown from 47% to 25% (Yes, at the expense of overall total return).

As well, I also added a code snippet to clean up all the delisted/'dead' positions that seem to accumulate in the portfolio over time, which really affected how orders were placed/$ allocated in the latter years of the backtest (I applied the methodology discussed in this thread: https://www.quantopian.com/posts/why-does-the-number-of-positions-and-leverage-creep-up-for-this-algorithm ).

Just wanted to share it here, as well as its accompanying tearsheet.

The tearsheet in the Q Research Environment

Hi can someone please convert this to the quantopian v2 syntax? I am not advanced enough to make the necessary changes yet.

Hi.

How did you calculate -0.6745? You said that it is an approximation for the top three-quarters of the market, but that doesnt make much sense to me.

Thanks

Does anyone have an update on this for 2017? Has performance still been consistent with the original?

This algo isn't exactly as the original because of the lack of historical annual fundamental data on Quantopian. It also uses a lot of leverage and this is bad. It was one of my first algo on Q. I'm going to reimplement it correctly not yet Q allows historical fundamentals with the Pipeline API.

It does

Following up here, has anybody updated this to work with the latest version of Quantopian syntax?