Mean reversion: how to trade a stretched price, and where it fails

Mean reversion is the tendency of a price that has moved unusually far from its average to drift back towards it. Traders measure how far price has stretched with Bollinger Bands, a z-score or RSI, then trade the move back to the average. It works in ranges and loses most when a market starts to trend.

In short

• Mean reversion trading bets that a stretched price returns towards an average, usually a 20-period moving average or, for day trades, VWAP (the volume-weighted average price of the session).
• A z-score measures the stretch: (price minus the moving average) ÷ the standard deviation. A close on the lower Bollinger Band, at its default setting of 20 periods and 2 standard deviations, is a z-score of -2.
• The target is the average, and the average moves towards price, so winners are usually smaller than losers. At 0.76R per win you need about 57% winners to break even before costs.
• Mean reversion fails in trends. Price can close outside the band bar after bar, and every add to a losing position makes the open loss larger.
• On a $25,000 Direct account the daily loss limit is $1,000, measured on equity, so open losses alone can end the account before a single trade closes.

What is mean reversion in trading?

Mean reversion in trading means selling a market after an unusually large rise, or buying it after an unusually large fall, because you expect price to return towards its recent average. The target is the average itself, so each trade aims for the snap back and exits there.

The "mean" is whatever average you measure from. Most traders use a 20-period simple moving average (SMA), day traders often use VWAP, and range traders use the middle of the range. Pairs traders use the normal price gap between two related markets, such as two share CFDs in the same sector.

The idea rests on overreaction. A run of triggered stop-losses or one large order can push price further than any change in value justifies, and when that pressure fades, price drifts back. Reversion to the mean can also happen with little movement in price: if price goes sideways at the extreme, the moving average falls to meet it, and your target moves closer to your entry.

How do you measure when price is stretched?

You measure a stretch as the distance between price and its average in units of normal movement, so that "far" means the same in quiet and busy markets. The usual mean reversion indicators are Bollinger Bands, a z-score of the distance from a moving average, and RSI extremes. Bands and z-scores measure the same thing.

Bollinger Bands

John Bollinger's bands sit 2 standard deviations above and below a 20-period SMA. The standard deviation measures how widely the last 20 closes are spread around their average, so the bands widen in busy markets and narrow in quiet ones.

The bands are not a 95% boundary. That figure assumes a bell-shaped distribution, and price changes have fatter tails, so closes outside the bands are more common than the textbook rate and bunch together in trends.

The z-score of distance from a moving average

The z-score puts a number on the stretch: z = (close minus SMA) ÷ standard deviation. With a 20-period SMA at 1.0900, a standard deviation of 15 pips (0.0015) and a close at 1.0862, z = (1.0862 - 1.0900) ÷ 0.0015 = -2.53.

The lower band is z = -2, so the z-score shows how far beyond the band price closed. Because it is a pure number, one rule, such as entering beyond -2 and taking profit at 0, works on any market or timeframe. A threshold of -2.5 or -3 gives fewer signals from more stretched prices.

RSI extremes

RSI(14) below 30 or above 70 is the classic stretched reading (our divergence trading guide shows the calculation). Some mean reversion traders use a 2-period RSI with thresholds of 10 and 90, which reacts within a few bars. RSI measures how one-sided recent bars were and ignores the average, and in a strong trend RSI(14) can stay above 70 for weeks.

Why does mean reversion work in ranges and fail in trends?

In a range, buyers and sellers broadly agree on value, so a move away from the average runs out of orders and comes back. In a trend, the agreed value itself is moving. Price can close beyond the band bar after bar while the average follows it down.

Traders call this walking the band. When a trend starts, the standard deviation widens, so the band moves away as price falls. The z-score can sit near -2 for a dozen bars while price drops 100 pips, and every one of those bars looks like a buy signal.

Four checks help you tell a range from a trend before you trade:

• The slope of the average. A 20 or 50-period SMA that is flat across the last few days suggests a range. One that is falling steeply says the "mean" is on the move.
• ADX (the average directional index), which measures trend strength from 0 to 100. Many traders treat a 14-period ADX below 20 as a range and above 25 as a trend.
• The higher timeframe. A 1-hour stretch inside a 4-hour range with tested support and resistance is a stronger candidate than one in the middle of a 4-hour downtrend.
• The economic calendar. A stretch caused by a central bank decision may be the first bar of a new trend, so check for scheduled news before you fade a move.

Where do you put the stop on a mean reversion trade?

Put the stop where the range idea is proven wrong: beyond the edge of the range or the most recent extreme, plus a buffer. Every mean reversion strategy buys weakness and sells strength, so every trade opens against the most recent move. Without a stop, a losing trade keeps growing for as long as it stays open.

That makes mean reversion the strategy most exposed to running losses. A price at z = -2 can go to -3 or -4 before it turns, if it turns at all, and buying more at the better price raises your position size just as the evidence turns against you.

The payoff shape adds to the pressure. Your target is the average, and the stop beyond the range edge is usually further away, so the reward is often smaller than the risk. Your break-even win rate is 1 ÷ (1 + R), where R is the reward per unit of risk: 50% at 1:1, 57.1% at 1:0.75 and 66.7% at 1:0.5. These figures are before costs, and our risk-reward ratio guide shows how costs push them higher. Small winners leave little margin: at 0.5R per win, one full 1R loss cancels two winners, and a 3R loss from an average-down cancels six.

Many traders add a time stop: close the trade if price has not reached the average within, say, 10 to 15 bars. A stalled trade still uses up part of your daily room, and a range that does not snap back may already be changing.

Worked example: a mean reversion trade on EUR/USD

This mean reversion trading strategy example uses a $25,000 Direct account, a plan to risk 0.5% ($125) per trade, and a 1-hour EUR/USD chart. Prices are for illustration. A pip in EUR/USD is 0.0001, and one standard lot moves $10 per pip.

1. EUR/USD has held between 1.0850 and 1.0950 for two weeks. The 20-period SMA is flat at 1.0900, and ADX(14) reads 16.
2. The standard deviation of the last 20 closes is 15 pips, so the bands sit at 1.0870 and 1.0930.
3. A sell-off closes a candle at 1.0862: a z-score of -2.53, with RSI(14) at 27. You wait.
4. The next candle closes back inside the band at 1.0874. You buy at 1.0874.
5. The stop goes 10 pips below range support, at 1.0840. That is 34 pips of risk.
6. Size: $125 ÷ 34 pips = $3.68 per pip, which you round down to 0.36 lots ($3.60 per pip). The risk is 34 × $3.60 = $122.40.
7. The target is the 20-period SMA. At entry it is 1.0900, 26 pips away: 26 × $3.60 = $93.60, or 0.76R. The break-even win rate at that ratio is 1 ÷ 1.76, about 57%.
8. Price drifts up slowly. Six hours later the SMA has fallen to 1.0893, because the low closes are now part of the average, and price touches it. You close at 1.0893 for 19 pips: 19 × $3.60 = $68.40, or 0.56R.

The mean came to meet price, and your reward shrank from 0.76R to 0.56R. At 0.56R, the break-even win rate is 1 ÷ 1.56, about 64%. That realised figure is the one to log in your journal, since it is what your win rate has to beat.

Costs reduce small targets the most. Simulated trades carry trading costs, as they would on a live account: a commission when a trade opens and when it closes, a price markup on some symbols, and holding costs on positions kept overnight.

How mean reversion works on CMC Funded Direct and Classic

On CMC Funded, the daily loss limit is a share of each day's starting equity, measured on equity, so the running loss on an open trade counts against it before you close anything. On the first day of the $25,000 Direct account above, the limit is $1,000; on a $25,000 Classic account it is $1,250. Reaching it ends the account, with no warning stage.

With a stop, the trade from the worked example can lose $122.40. Eight full stop-outs ($979.20) fit inside the Direct limit, and ten ($1,224) inside the Classic one.

Now take the same idea without a stop. You buy 0.36 lots at 1.0874, add 0.36 lots at the next band touch at 1.0856, and add again at 1.0840 when price breaks the range. A data release then sends EUR/USD lower.

On Direct the account ends at 1.0764 though none of the three trades has closed. With one position and a stop at 1.0840, the same move cost $122.40.

The maximum loss is the second limit to plan for. It is fixed, so the floor never moves: $23,500 on the Direct account and $22,500 on Classic. With no gains yet, that is room for twelve $122.40 losses on Direct ($1,468.80) and twenty on Classic ($2,448). Full limits for every size are on the rules page, and the daily loss and maximum loss lesson shows how to work out each day's room.

Count trades on related markets as one: buying EUR/USD and GBP/USD at their lower bands during a dollar rally is the same bet twice, and both open losses count against the same daily limit. Set a maximum number of attempts per day before the session opens, too.

Is mean reversion the opposite of trend following?

Yes, in the bet it makes: mean reversion expects a stretched price to come back to its average, while trend following expects a price that is already moving to keep going. The two do well in opposite market conditions. Some traders run both on different markets or timeframes and let a filter such as ADX decide which one applies.

Common mistakes

• Buying the first band touch in a downtrend. Check the slope of the average and ADX before you treat a stretch as a buy.
• Averaging down with no total limit. Each add raises the open loss for the same move. Decide the total risk for the idea before the first entry, and put one stop on all of it.
• Aiming for the opposite band. The far band is a trend-sized move from the lower band. Take profit at the average.
• Treating RSI below 30 as a signal on its own. In a strong trend RSI can stay low while price keeps falling. Combine it with a range and a trigger, such as a close back inside the band.
• Moving the stop further away because the z-score has reached -3. The stop marked the price where the range idea fails, and that price has been reached.

Questions traders ask

What is the difference between mean reversion and regression to the mean?

Regression to the mean is a statistical effect: an unusually extreme measurement tends to be followed by one closer to average, because chance contributed to the extreme. Mean reversion in markets is the idea that prices drift back towards an average over time. A mean reversion trade also needs that average to hold steady while you wait.

How do you test whether a market is mean-reverting?

Quantitative traders run statistical tests on the price series. The augmented Dickey-Fuller test checks whether a series tends to return to its mean, and a Hurst exponent below 0.5 points to mean-reverting behaviour. A chart trader can log 50 band touches and count how many reached the average before the stop.

How long does mean reversion take?

It depends on the chart. A 20-period average on a 5-minute chart covers 100 minutes of trading, so trades there last minutes to hours. On a daily chart the same average covers about four weeks, and trades last days or weeks. Set a time stop in bars so you know when the idea has expired.

What is VWAP mean reversion?

VWAP is the average traded price of the session, weighted by volume. Intraday traders treat the distance from VWAP as the stretch and VWAP itself as the target. It resets every session, so it suits day trades. On forex, VWAP uses tick volume, because currency trading has no central volume record.

Next steps

The mean reversion and range trading lesson walks through spotting a range one step at a time. Before each trade, the position size calculator turns your stop distance into a lot size you can afford under the daily limit.

You can compare Classic and Direct on the challenges page.