In my search for probabilistic strategies I like to start from very simple ideas. The kind of ideas that anybody can implement without complicated tools and without a PhD. This article describes such a strategy, and shares basic points about choosing between variants.
The idea here is to perform a weekly rotation of index ETFs based on their momentums.
I have chosen 5 indexes:
The first rule is, once a week, to identify the ETF which has the highest relative increase in price for 1 year, and to take or keep a position on it, after selling the previous position if necessary. For every rotation, the capital plus gains or minus losses, plus dividends, are reinvested.
You can pick the right ETF using Yahoo, Freestockcharts, or any online charting tool that allows you to compare stocks and ETFs. You just have to put the 5 usual suspects on the stage once a week, click on the "1 year" button and select the highest percentage on the current date.
The following chart simulates the described weekly rotation from January 2002 to May 2012 (return in red, S&P 500 in blue).
The second idea is to add a market timing rule.
First this one: Before choosing the ETF every week, calculate the 5-day and the 20-day simple moving averages of the S&P 500 (SMA5 and SMA20). If SMA5<=SMA20, sell if necessary and stay out of the market.
In fact you don't have to calculate anything. Simply add both indicators on a daily S&P 500 chart and compare them:
This is not a very good market timing indicator, but it helps to keep things simple and this article short. Eventually a simulation shows that it works quite well here (return in red, S&P 500 in blue):
As a second variant, I propose to test a calendar based timing.
Below is a simulation chart without the moving averages rule, simply staying out of the market when the rotation date is in January, February, May, June, July, August or September. I have explained in a previous article why I think that these months have statistically a negative effect.
The total return is 243% with only 5 months of exposition per year.
I propose now to combine both calendar and moving averages rules. Here is the result of a simulation since 2002:
The return is 161%. Lower than the calendar-only variant, slightly better than the SMA-only one. Two advantages are a shorter exposition in time and a protection if seasonal effects don't work as usual.
Now I would like to share some of the criteria I use to select or reject a strategy. This is not only about return, but also about drawdown and probabilistic robustness.
Here are some indicators about the strategy with the SMA market timing ("SMA"), with the calendar timing ("cal"), and with both.
AR: average annualized return in %
DDM: max drawdown in %
DDL: max length of drawdown in weeks
K: a probabilistic robustness indicator (which is Kelly ratio in %)
Adding a calendar timing not only increases the return, but also lowers the drawdown's maximum depth and maximum length, and improves significantly the robustness indicator.
The highest return is obtained with the calendar timing alone.
The lowest drawdown, highest robustness, and shortest exposition is obtained combining both rules. Nevertheless I don't consider it as a very good strategy because DDM is higher than the average return.
As a conclusion, it is possible for an individual investor to execute strategies and outperform the market without complicated tools. However, this is only a starting point. It takes much more work to identify, test, select, understand and finally decide to put money on really valuable strategies.