How To Develop a Trading System Part 3

See other posts in this series…
[catlist id=2 numberposts=3 pagination=yes]
ABSTRACT When most people embark on a journey, they get a map, they consult an expert, or at least someone that’s been there before, and developed some reasonable outlook. The problem with trading, is that most people think it’s about banking huge profits, taking huge calculated risks, and living the life style of the rich and famous. That’s not it at all. Good trading is routine and can be boring, that is if t’s done right…and that’s what today’s lesson is about, a boring but CRUCIAL concept called Correlation.

Correlation Coefficient

In the previous lesson (Part 2) we learned that strong strategies are ones that exhibit anti-fragile behavior, and are made up of multiple non-correlated strategies. So, what does it mean to be non-correlated?

Let’s first discuss what correlation actually means. It’s a statistical property between 2 things that measures the strength and direction of the relationship between those two things. The relationship can be visualized on a scatter plot as how closely data, that is generated by the two things, comes in line with one another, and is measured as a number between -1 and 1.

This is a plot of trade results by 4 different pairs of trading strategies, mapped with standard run of the mill scatter plots. If you took a statistics course in school, I’m sure you ran across one of these babies.


Plot (a) has a tight correlation with a value of +1.0, notice also the direction of the best fit line, from bottom left to top right. Plot (b) is a correlation of -0.5, (c) is +0.85, and (d) is +0.15.

From a human point of view, +1.0 is perfectly correlated, and plot (c) at +0.85 is considered to be strongly correlated. The other plots simply aren’t in the same ball park.

When we analyze whether two or more strategies are correlated, it’s obvious that we can’t simply look at them and see the correlation. You need special tools to do that. Fortunately those tools are available. You can however make a pretty good guess as to which strategies exhibit a high degree of correlation, and which ones are more likely to not be correlated.

For example, if you had two strategies that were both based on momentum indicators, that determined their respective values by evaluating price with an overbought and oversold zero-based oscillator, like RSI and Stochastic, then you can reasonably assume these two strategies are probably highly correlated. And they are, because both are based on price action, and both measure overbought and oversold.

Highly Correlated Strategies Are Undesirable

Remember the principle of having multiple non-correlated strategies? The first part is “multiple,” you need more than one strategy in your portfolio, three is better, four is slightly better, having five or more starts to get hard to measure how much better, better is.

The 2nd part of the principle, is that the strategies are non-correlated, as measured from a statistical point of view. A correlation of between +0.5 and -0.5 is desirable. If you can get the range even tighter or closer to 0.0 that’s even better. If I saw a correlation like in plot (d), that would make me very happy, plot (b) would also make me happy, nut less so. Plot (b) would represent my limit. I wouldn’t want strategies to be any more correlated than that.

Highly correlated strategies will react the same way to similar market conditions. So if the market suddenly drops, and your strategies are highly correlated, and they are on the wrong side of the trade—which can happen—then all your strategies are going to experience draw down. And that is bad. If however, you have non-correlated strategies, then the odds are that some of those strategies are going to be on the opposite side of the trade, prroviding a natural hedge, and protection against such occurrences.

This is why you need multiple strategies, so that you can have enough strategies that might be hedging a bad choice in direction. But you may think, wouldn’t that work in the positive direction as well. The answer is yes, however if all the strategies have a bias towards being successful—why else would you trade them— then over all you will see positive results.

Strategies that are run simultaneously, with very little correlation to one another, will have a much better chance at handling the infinite variety of market conditions that exist. And these non-correlated strategies, although individually may not always look that great, when combined, they look unbelievably good. And that’s all due to the anti fragile nature of multiple, non-correlated strategies.

I hope this makes sense to you, because it is the principle reason why our auto traders work so well in virtually any market condition, including violent market corrections, price upheavals, price shocks, flash crashes, and even simple choppy or trendy markets. It doesn’t matter, the combination of diverse strategies are fare more powerful than any of them individually.

How Long Does a Correlation or Non-Correlation Last?

This is the big question, because the answer is…I have no friggin idea, nobody does. Some relationships between strategies last a very long time, those are the ones we like to gravitate towards, some don’t last very long at all…those are the one we tend to avoid. There are some definite clues as to which strategies will last, and which will not, the rest is experience.

But none of that is a guarantee. That’s why we must always be developing and testing and evaluating new strategies for possible promotion into our portfolio. This is why we run something I call campaigns, that typically last about 3 months, or one calendar quarter, and reevaluate the portfolio of strategies, and compare it to other potential portfolio mixes.

Strategy Farm System

So, this is the methodology part of the How to Develop Trading Systems. And quite frankly it’s the fun part, if you’re into finding new ways to trade, and evaluating new strategies as potential members in your farm system.

I’ve made the analogy that the farm system is just like the farm system of a major league baseball team, where the major league team is our working portfolio, and the minor league team(s) are the farm system of strategies that we are trying to develop, incubate and cultivate for potential promotion onto the major league team.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.