The examples presented on these pages are for educational purposes only and may not
be seen as investment advice! You, and only you, are responsible for what you are doing with your money.

Please click here to read the entire disclaimer and disclosure.

The goal with this strategy is to exploit deviations from the mean on a timeseries consisting
of the two german utilities RWE and E.ON, while being neutral to overall marketchanges in the utility sector.

To find out wether this pair is appropriate for this type of trading,
I made a linear regression analysis, as shown in the graph below.


The equation shown in the graph is the equation for the theoretical price of 1 piece RWE:
RWE = 2,2298 * E.ON + 1,3152 EUR.

This equation explains 88,41% of all the prices of RWE in the period examined, shown by the R sqared.

1,3152 EUR is what an investor on average has to invest to own a portfolio of either 1 long RWE
and short 2,2298 E.ON and the other way round.

The next step is to calculate all the theoretical prices of RWE and to divide these theoretical prices of RWE
by the observered prices. This generates the ratio, where a ratio of 1 means that the prices are in equilibrium,
while a ratio smaller than 1 means that E.ON is cheap relative to RWE and greater than 1 means that RWE
is to cheap relatively to E.ON.

Plotting the ratio, as shown below, one can easily see that the ratio is mean-reverting.


So according to the above a simple strategy could be to buy 2,2298 pieces of E.ON and to sell short
1 piece of RWE when the ratio is below 1 and the other way round when greater than 1.

To find a suitable entry and exit price I calculated the standard deviation of the ratio
and backtestet the strategy with 1, 2, 3 and 4 standard deviations from 1 as the entry point.
The trades will be closed when the ratio moves back to 1. A stop-loss is not applied.


As the graph above shows, 1 standard deviation from the mean seems to be the best entry point.
Admittedly this is before trading costs and slippage.

One might have wondered why I only examined the data until end of 2006 in my regression.
The answer is easy: to have enough data to backtest "out of sample", meaning that the model
is not fitted to this period and therewith kind of a "live test".

The graph below shows the performance if one had made this model on New Years eve 2006 and
had started trading it in the beginning of 2007:

Out of Sample Backtest

The beta of this strategy is fairly close to zero (0,004), meaning that the strategy is close to marketsneutral.
The Sharpe Ratio is 0,48 (assuming a risk free interest rate of 2%).