Mean reversion: buying weakness, selling strength
What a mean reversion system does, what its rules tend to look like, the shape of its results, and the conditions where it struggles.
By Baris Demirci · · Rev00
The idea
Prices rarely move in a straight line. They stretch away from a typical level, then often come back toward it, like a rubber band. A mean reversion system is built around that snap-back. It buys after a fall, on the view that the fall has gone further than it should, and sells once price has come back.
It is one of the two oldest families of trading systems. The other is trend following, which does nearly the opposite.
What the rules look like
Almost every mean reversion system has three parts:
- A measure of normal. Usually a moving average, or the recent range of prices.
- A measure of stretched. How far price has fallen below normal. It might be the distance from the average, several down days in a row, a close near the bottom of the day’s range, or an oscillator built for the purpose.
- An exit when price comes back. Back to the average, a bounce of a set size, or simply a set number of days.
Many versions add a filter, such as buying dips only while the longer-term direction is up.
The shape of the results
Mean reversion has a recognisable pattern:
- It is right often. Most stretches really do snap back, so most trades end in a small gain.
- The gains are small. It sells as soon as price is back to normal, so it never rides a big move.
- The losses are rare but large. Sometimes a fall is not a stretch at all. It is the start of a much bigger move down, and the system bought the first step of it.
That shape feels comfortable most of the time, which is exactly what makes the rare large loss so dangerous. A backtest that happens to miss the bad periods makes the system look far safer than it is.
Where it works, and where it struggles
- It does well in choppy markets, where prices swing back and forth without going far.
- It struggles in strong, sustained moves. In a falling market it keeps buying, and each buy is followed by more falling. The old warning about “catching a falling knife” is about exactly this.
- It trades often, so costs matter. In thinly traded markets, the spread can take most of what each trade is worth.
- It depends on the market and the horizon. Reversion shows up in some markets and at some time scales and not others. Mean reversion is not a law of markets goes into that in detail.
What it asks of you
- Buying when it feels wrong. The system buys after bad days, often when the news is at its worst.
- A hard decision about stops. A tight stop sells exactly the kind of dip the system is built to buy. No stop at all leaves the rare large loss fully open. There is no free answer. Decide it in advance and test it.
- Respecting the rare loss. Size each trade so that the large loss, when it comes, is one you can live through and keep following the rules.
The short version
A mean reversion system buys what has fallen too far and sells it when it comes back. It is right often, wins small and loses rarely but large. It does well when markets swing and badly when they trend, and its worst moments come from treating the start of a big move as a dip.
Written for people building their own systematic portfolio management process. Nothing here is a recommendation to buy or sell anything. See the disclaimer.