A reader of my book sent me a question that combines system design and testing issues with risk management considerations.

Hi John,

I backtested a strategy on all S&P 500 plus Nasdaq 100 stocks. After checking for outliers, I selected the 10 best symbols in terms of risk adjusted returns and statistical significance. My idea is to trade 10 stocks to increase the number of trading opportunities in a trend following system. Following your book’s approach, I checked fixed vs variable vs stepped sizing methods and determined that in all cases the variable method was the better one. So I ran the calculations and the comparisons again using variable sizing methods, and finally came up with optimal fixed fractions of equity to trade for each stock, given acceptable drawdown limits.

So now I have the following problem: I have different fractions for each stock but one trading account. I am thinking of allocating one tenth of the trading account to each stock and applying the stock’s fraction to that amount, but I am not sure if it’s the right approach.

Thank you !!

Rod

First, I’d like to congratulate Rod on going through the full design and testing system development process. I will tell you flat out that my work doing these sorts of things early in my trading development helped me enormously to understand how different indicators and systems methodologies work. That, in turn, helped me work through the process of finding my trading niche.

Now, addressing Rod’s question….

There are a couple of things I would suggest need to be looked at and thought about at this point in the process. While it’s admirable that Rod is looking to track a specific group of stocks to provide the frequency of trading opportunity he’s looking for, he needs to consider the risk of all 10 stocks moving against him at the same time. My guess is that the stocks he’s selected are relatively strongly correlated, though that’s worth testing. It means they could all move against him at the same time, which introduces the risk for a substantial loss at some point.

That said, I think Rod would do well to perform an additional set of tests. It sounds like he’s done singular system tests – meaning testing the system’s performance on an individual security basis. It does not, however, sound like he’s tested the system in a unified way. By that I mean running a test which encompasses all the 10 securities he’s selected as a full portfolio look. That sort of thing will help identify the risks of correlations and see the kinds of drawdowns it could produce (potentially), and let him determine his position sizing in aggregate rather than just security by security.

Look at the characteristics

The last bit of advice I’d offer Rod at this stage is to try to identify the set of charactersistics which makes those 10 stocks he’s selected good for his system. Individual stocks will change how they trade at different points – sometimes temporarily, sometimes permanently. If you know what makes a stock good for your system you can track the ones you are using to see if they change, and you can keep an eye out for other stocks that are good candidates for inclusion.

Struggling with support & resistance and knowing what the key market levels are? Check out the Price Distribution Analysis methods I use.

Bookmark to:

Risking Something Other Than 1-2%
Money Management Question
How do you set position sizing in Metastock system testing?
Fixed Income and Stock Trading
Decembers Most Delightful