Building a Trading System: Part 2 - Mean Reversion

To start out our trading system we need a theory. We need an idea that makes sense that will tell us when to buy and sell. Mean Reversion is one such theory. I didn't invent it, and in fact it's simply a mathematical model that is frequently applied to finance, or in this case, to trading. To give you some idea of what mean reversion is and how it's used in financial model check out this Wikipedia article. There is also a simple description on Investopedia.

For the purpose of developing a trading system, let me take a stab at an explanation. Mean reversion is the theory that stocks trade around a mean. Sometimes they get priced too far away from that mean for one reason or another, and when they do they have a tendency to revert or return to the mean. Let me show you an example:

What you see here is a chart of daily stock prices. In this case it happens to be the SPY ETF, but it could be any stock, index, commodity or almost any other pattern of prices. The thin RED line happens to be a 5-day simple moving average. This is what we will call our mean for this example. It's simply the average of the closing price for the last 5 days. It could be many things... this is just one example.

You'll notice that I've placed red and green arrows where the price has moved far away from the red line. In both cases the price moved back to the red line within a couple of days. Does it always happen this way? Absolutely not, but our theory is that this is a tendency, and that theory is what we'll build our trading system on. If it's wrong (or right) we'll find out when we backtest it.

This is the simple theory: When stock prices are stretched far away from the mean, they have a tendency to revert to it. It's a little like the sling shot I played with as a kid. When I pulled back and let go of the pouch connected to the rubber bands, it shot back toward the handle. And more interesting, the further I pulled it back, the farther the ball bearing flew (ball bearings are much better than rocks). Our theory might be tested for this same concept. Now that you have an idea of what Mean Reversion is, there are a heap of questions to be answered:

• What's the best way to measure the mean?
• How far is far away from the mean?
• How far back toward the mean do we expect it to move?
• How long does it need to stay away from the mean?
• How can I turn this information into a trading decision?
• What stocks work best with this theory?

And those are just a few. There are no simple answers, and in fact most answers are "it depends". There are many ways to build a mean reversion trading system, so what you will see is just one. The goal is to share enough information that it will be obvious how many different approaches can be taken, and how they can be modified for different results.

Good Trading...