Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
9000+ Securities Momentum based strategy - 1 mo holding period

Hello Guys,

I wrote an algorithm that analyzes over 9.000 (from penny to blue-chips) US stocks over some metrics and recommends wether to buy or not based on a series of inputs and data (i've come to understand that momentum is the main driver for predicting the next month's return for any given stock).

The algorithm uses machine learning for selecting the individual stocks and it is a long-only strategy that uses momentum as the main driver for looking for hot-stocks with high bullish probability. The portfolio has monthly rebalance and can completely change the portfolio composition between on month and another (even though that's not the case most of the times).

In order to allocate the capital between each picked stock, and also between bonds (for turbulent market periods) and stocks (for less risky market periods), I used the Modern Portfolio Theory (MPT) developed by Markowitz and further developed by a number of other scholars in more recent years.

The results are promising. I've back-tested the algorithm since 1990 with average 22% annualized returns, 26% maximum drawdown and 0.90 Sharpe Ratio over a 26 years period (as a comparison case, Buffet's Berkshire Hathaway has 0,76 Sharpe Ratio over a 30 years period).

It has been tested out-of-the-sample since the beginning of the year, and it now has 4.25% (looks fairly great if compared to S&P index 0.75% return over the same period).

I've made a website to share my progress with the Algorithm, in which I'll share (for free) the recommendations it gives for May/2016 - for people that are interested to learn a little bit more about it - on the newsletter in the website:

http://stockreturn.strikingly.com/

If you have any further questions or comments on it, please, let me know. It'd be an honor to talk a little bit more about this algo with you guys.

PS: I'm not sharing the graph here because it is not possible to analyze a universe of stocks of that size here in Quantopian. I'm just posting it here to show you guys that there may be a great opportunity on analyzing a greater volume of data and different securities when developing a trading strategy.