First portfolio
In previous articles, we created and tested three types of strategies and their variants: Simple breakout strategy, Breakout limit strategy, and Buying dips strategy.
Today we will create our first small portfolio from them.
Regardless of what our individual strategies look like – whether they have a perfect equity curve or not – our future trading success will never depend on the performance of individual strategies, but on the overall portfolio of many uncorrelated strategies. As I once read, the only “holy grail” in trading is the diversification of strategies in the portfolio.
Only a portfolio composed of different strategies will protect your investments even in periods when individual strategies fail. A portfolio is like a hockey team at the World Cup – individuals can be great, but if they don’t play together as a team, they have no chance of winning.
Risk normalization
First, we test all strategies again with a different money management. The aim is to normalize the risk so that all strategies have the same starting position.
As a reminder: we have 10 strategies with different logic and different outputs – namely Breakout (4x), BreakoutLimit (3x) a BuyDip (3x). The total starting capital is $100,000, with each strategy being assigned a risk of 10% of the account. We then retest the strategies one by one.
Creating a gross portfolio
Once we have the strategies tested, we create a rough portfolio to visually assess the similarity of the strategies and calculate their correlation.
We look at the correlation between the strategies on a monthly basis (Correlation by Month) and separately analyze their correlation at losses (Correlation by Loss). I only use Loss because I don’t mind similar gains, but I don’t want to have similar losses.
The correlated table looks like this:
Portfolio optimization
From the correlation table we can see that there is a high correlation between the strategy Breakout_var03, Breakout_var04, BreakoutLimit_var03A a BreakoutLimit_var03B. It is quite logical, because for these two Breakout strategies we only changed the number of traded shares and for the BreakouLimit strategies we only changed the entry price.
So we eliminate two strategies – e.g. Breakout_var03 a BreakoutLimit_var03A – and we will create a new portfolio.
Here is the new correlation table.
Upon further analysis, we find that there is still a high correlation between BreakoutLimit_var03B a Breakout_var04. Since we have more breakout strategies in the portfolio, we will remove Breakout_var04.
We will create a new portfolio.
The table already looks beautiful green. Perhaps we could still eliminate one BuyDip type strategy, where we also have a correlation higher than 0.3. Finally, we will try to eliminate the strategy BuyDip_var02.
We will create the last 4th portfolio.
In the end, we eliminated four of the original 10 strategies and created a final portfolio with low correlations. The graph of the individual strategies in the portfolio looks as follows.
Statistics of final portfolio.
Benchmark comparison
The final comparison of the portfolio with the benchmark shows a clear difference between passive investing (holding SPY for 40 years) and active trading. The table on the right shows the statistics of both the portfolio and the benchmark at the same % drawdown – and it can be seen that our portfolio clearly wins.
Summary
Building a portfolio is like selecting players for the World Cup. The best strikers, defenders and great goalkeepers must be selected. Everyone has a different role in the team, but overall they have to form a strong invincible team.
And it is the same with the selection of strategies for the portfolio. We choose a different strategy for each market deviation so that the overall portfolio is strong, robust, and resistant to large market fluctuations.
We have completed the basic structure of the portfolio. Next time we will focus on fine-tuning risk management and deploying the portfolio to the test account.
Libor Štěpán
