A "Markowitz" Attempt
As posted in a Quantopian forum.
This is a peculiar trading strategy. The original author probably wanted it to be based on some Markowitz portfolio management principle, but it is not. Nonetheless, over part of its trading interval, it does make as much money doing nothing as it does trading.
I have not found why over some time intervals it does not trade at all. For instance, prior to 2007 no trades, and just after October 2017, again no trades. I limited the simulations from 2009/01/02 to 2018/09/21, even if it exceeded the October 2017 no trading period.
Whatever the stock inventory was after October 2017, we could say the strategy went into hibernation or in a Buy & Hold scenario until termination time. From the charts below, it turned out to be quite profitable.
I usually get interested in trading strategies that can withstand 10 years and more. This one with 9 years barely qualifies. However, I was interested because its optimizer was the same as the one used in the first chart in this thread.
I added many changes to the original design giving it a totally different long-term outlook. Applying pressure where I thought it would add value. Thereby, changing the nature of the betting system implemented. I increased the number of stocks to 28. And since the strategy was scalable, all trading profits were reinvested as the strategy progressed in time. I even used some leverage.
I have not as yet added protective measures. They should come next if I find the time. I consider the strategy as still in its development stage.
Here are the chart, first the cumulative returns:
Some of the metrics:
And summary stats:
The strategy made 2,968 trades and manage to take its $10M initial stake up to $1.1B, a 63.8% CAGR over the 9 years.
I should be able to add protective measures that should increase overall performance as well as reduce drawdown. It is a pity I do not know why it stops trading in October 2007. However, if it was some added feature, base on the results, I would be ready to keep it on the condition that it also works on other stock selection methods.
Prior to 2018, the strategy seemed ready to break down. Yet, letting it do whatever it did which was nothing at all, it proved a lot more lucrative.