Mean reversion trading strategy: fading stretches back to the average

What mean reversion is

Mean reversion is the idea that prices move around an average and that extreme moves away from it tend to be followed by moves back. A trend follower sees a strong rally and buys, expecting more. A mean reversion trader sees the same rally, notes how far it has gone from the average, and sells, expecting some of it to unwind.

Both are right at different times. Markets spend long stretches going sideways, especially major currency pairs between big news events, and in those periods stretches away from the average do tend to fade. Then a trend starts and the same stretch becomes the first leg of a much bigger move. So a mean reversion strategy needs two parts: a way to measure the stretch, and a filter that switches the strategy off when the market is trending.

The average can be a simple moving average, the VWAP for intraday traders, or the middle of an established range. This page uses moving averages and Bollinger Bands, because they measure both the average and how unusual a move is, in one tool.

The regime filter: is the market ranging?

Check this before anything else, on the 4-hour chart. You want at least two of these three:

• A flat 50-period moving average. If it has moved less than about 30 pips over the last 20 candles on a major pair, the market is not trending in any meaningful way.
• Price crossing the average repeatedly. Several crosses in the last week means price is oscillating around it, not leaving it behind.
• ADX below 20. The Average Directional Index measures trend strength regardless of direction. Low readings mean no strong trend.

Then check the calendar. A central bank decision, an inflation figure or a jobs report can turn a quiet range into a trend in one candle. No new mean reversion trades in the 12 hours before a high-impact release for either currency in the pair. Our page on how forex prices move explains which releases matter.

Some pairs range more often than others because the two economies move closely together: EUR/GBP and AUD/NZD are the classic examples. They are good markets to test this strategy on first.

The setup: outside the band, then back in

On the 1-hour chart, add 20-period Bollinger Bands at 2 standard deviations and a 14-period RSI. The middle band is the 20-period simple moving average; the outer bands mark moves that are large compared with recent volatility.

1. The stretch

A 1-hour candle closes outside the lower band (for a long) or the upper band (for a short), and RSI is below 30 or above 70. Two independent measures agree that the move is unusual.

2. The snap back

The next candle, or one soon after, closes back inside the band. This is the entry. Buying the first close outside the band means catching a falling knife; waiting for the close back inside means the move has at least paused.

In the chart, EUR/USD has been ranging around 1.0850 for days. A sharp drop closes below the lower band with RSI under 30, low at 1.0815. The next candle closes back inside at 1.0828, which is the entry. The stop goes at 1.0812, a few pips below the low, for 16 pips of risk. The target is the middle band, near 1.0850, 22 pips away.

Stop, size and exit

The stop goes a few pips beyond the extreme of the stretch: below the low for a long, above the high for a short. If price makes a new extreme after you have entered, the stretch has resumed and the reversion idea is wrong. Some traders use 1.5 times the 1-hour ATR from the entry instead; both are fine, as long as the stop is decided before the trade.

Size from the stop as always. With a $10,000 account and 1% risk you have $100; a 16-pip stop on EUR/USD costs about $160 per standard lot, so the position is 0.62 lots. The position size calculator will check it for you.

The exit is the middle band. Because the middle band moves, update the take-profit each hour, or simply close when a 1-hour candle touches it. Be honest about the arithmetic: here the reward is 22 pips against 16 of risk, about 1.4 to 1, which breaks even at about 42% before costs. Mean reversion trades often have smaller reward-to-risk ratios than trend trades, so costs and the occasional large loss matter more. Add a time stop: if price has not reached the middle band within 12 hours, close the trade, because a range that stops reverting is often about to stop ranging.

Never add to a losing mean reversion trade. The logic of "it is even more stretched now, so it is an even better trade" is exactly the logic of the martingale, and it works until the one trend that does not come back.

When mean reversion fails: the band walk

Mean reversion fails in a trend, and it fails in a particular way. In a strong trend, price does not touch the outer band and come back; it rides along it, closing outside or at the band candle after candle. Traders call this walking the band. RSI stays above 70 for days. Every short taken because the market looks overbought is stopped out, and each new one is taken at a higher price.

The gold chart shows a band walk. Price grinds higher along the upper band for two days, and RSI never falls below 60. A trader fading each push takes three losses. The regime filter would have prevented all three: the 50-period average was rising steeply and price had not crossed it in days.

The other failure is the news breakout. A range can look perfect right up to a release, and then price leaves it for good. This is why the rule against trading before high-impact news is part of the strategy, not an optional extra.

Finally, watch for a change in volatility. When the Bollinger Bands contract to their narrowest in weeks, a big move is often building. A squeeze is a reason to stop fading and wait to see which way the market breaks, not a reason to trade the edges more often.

How to backtest a mean reversion strategy

The rules are mechanical, so this strategy is well suited to a careful test:

1. Choose one pair, EUR/GBP or EUR/USD, and go back at least two years on the 1-hour chart, so the test includes both ranging and trending periods.
2. Apply the regime filter on the 4-hour chart for each week before you look at that week's 1-hour candles.
3. Record every close outside the band with RSI beyond 30 or 70, every close back inside, and the result at the middle band, the stop or the 12-hour time stop.
4. Charge the real spread on every trade, and record the trades the filter rejected, with what would have happened.

The key comparison is filtered against unfiltered. If the filter does not make a clear difference, it is not doing its job. Look at the largest single loss and the longest losing streak as well as the average result, because mean reversion strategies tend to look smooth until a trending month arrives. The drawdown calculator and our backtesting guide help you judge whether you could live with the worst stretch.