Didn't Markowitz win a nobel prize for the stuff, but ended up doing equal weights stocks and bonds for his own portfolio? or something like that...

Here are the results using the same assets, but without any charlatanism (science) added to it, just equal weighted, monthly rebalancing.

Why add mumbo jumbo when you can do better by being "naiive"?

Sorry, I kept too much cash in my other backtest, here is one with no cash...

Or since our universe only contains stocks we know have done well in the past we could just assign weights randomly.

James, so da*n funny...this entire industry is a joke.

Here is the same "algo" with 98% of cash invested at all times...now give me your 2/20 before I close my fund for new investors.

In closing, I am adding some leverage to the same "algo"...now turn over your entire life savings PLUS morrgage your house and invest it all in my black box fund, before I close it for new investors ;)

Hi James

To be clear, I did not cherry-pick these stocks. They are the exact same ones from the original post https://www.quantopian.com/posts/the-efficient-frontier-markowitz-portfolio-optimization-in-python-using-cvxopt. The authors of that post are: Dr. Thomas Starke, David Edwards, Dr. Thomas Wiecki...

My algo above doesn't actually start trading until 2007, I have corrected the code to make this clearer and will show the results in another reply later. Attached is YOUR random portfolio construction results (2007-2015) I got when I ran it.

Thanks

Peter

Hi Behavioral Trader

Well where do I start... First off there is no charlatanism, mumbo jumbo and the entire industry is not a joke. It would be a better community, willing to share more and help each other, if members did not jump to conclusions and make such remarks - such behaviour is not conducive.

My algo above doesn't actually start trading until 2007, I have corrected the code to make this clearer and will show the results in another reply later. Attached is YOUR equal weight portfolio construction results (2007-2015) I got when I ran it.

Thanks

Peter

To the rest of the Quantopian community who don't jump to conclusions...

Before I post the results of the toy example, let me clear the air:

• the intention of this post was to show that it is possible to use Markowitz style portfolio optimization in Quantopian to build portfolios that suit your return/risk preferences. On the other thread, there were quite a few unanswered questions on how to do it
  
• it was never about who has the best algo
  
• it's not the only way to construct portfolios, like I said from the get-go it's a toy example, but I personally think it's worth having a look at and experimenting with. There are a lot of tweaks you can do to improve it...
  
• of course, there's a lot more to know and you need to do your own homework!

Best of skill

Peter

One thing I want to highlight is that your estimates of the covariance matrix and the expected returns are key for better optimization results. The optimizer is very sensitive in particular to the expected return input.

I have just used some adjusted form of sample mean and sample covariance in the toy example, but this is where your skill in modelling should be applied...

The amendments I made to the original code are as follows:
- aligned the trading start date with the algo start date
- changed commission and slippage to the defaults
- changed rebalancing to time_rules.market_open(hours = 1, minutes = 30)

Conclusion: outperforms random and equal-weight construction techniques.

Also, note how the beta-to-SPY is low enough for the contest...

Thanks

Peter

Yes this industry is full of charlatans who preselect securities for their backtests, play with backtest windows, but do not even check if their algo trades, or if it trades all the right securities half of the time. Some of them even shamelessly plug their websites and other spam.

LOL. The internet is full of trolls making unfounded accusations and misrepresenting the facts.

Replaced the stocks with the largest market cap stocks at the beginning of the backtest.

Equally weighted has Meucci beat.