TradingView Black Friday: up to 80% off annual plans, 23 November to 1 December. See the dates and the best plan

Kagels Trading
Topics

Mean Reversion Trading: Rules, Indicators and a 33-Year SPY Test

Contents
  1. Mean reversion trading in 30 seconds
  2. What is mean reversion?
  3. The math behind it, for advanced readers
  4. How does mean reversion trading work?
  5. When does mean reversion work best?
  6. Mean reversion vs. trend following
  7. Five indicators for mean reversion
  8. Our test: two RSI rules on SPY since 1993
  9. A chart-based alternative: the 1-2-3 reversal
  10. Pairs trading: mean reversion between two markets
  11. What does the research say?
  12. Pros and cons of mean reversion trading
  13. Who is mean reversion for?
  14. Conclusion: is mean reversion the right strategy for you?
  15. Frequently asked questions about mean reversion trading
  16. About the author

Hardly any price moves in a straight line. After a strong move up or down, most markets are pulled back toward their average. That observation is the core of mean reversion trading: you buy when others sell in panic and sell when euphoria has pushed the price too far. It is one of the few approaches that deliberately trades against the current move.

This guide shows how the principle works, which indicators make the distance from the average visible and what two common rule sets did on SPY over 33 years. We ran the test ourselves, with the rules fixed before we looked at the results. The numbers are less flattering than many blog backtests, and that is the point: mean reversion delivers high hit rates, but hit rates alone do not make a strategy. This article follows our editorial policy.

Mean reversion trading in 30 seconds

  • Definition: mean reversion is the tendency of prices to move back toward their long-term average after a strong deviation.
  • The logic: you buy oversold markets and sell overbought ones, against the current move, expecting the correction.
  • Key tools: Bollinger Bands and Keltner Channels mark extreme zones, the RSI measures the overextension, and the z-score makes it comparable across markets.
  • Our test: two RSI rules on SPY won 78 to 80 percent of their trades from 1993 to 2026, but returned 1.7 and 5.1 percent a year against 10.4 percent for buy and hold.
  • The risk: in strong trends the average moves away, and the expected correction does not come.

What is mean reversion?

Mean reversion is the statistical tendency of a price to move back toward its historical average after a strong deviation. Traders use it to bet on corrections after overreactions. Extreme moves correct in most market phases, and that observation is the base of the approach.

An example makes it concrete. A stock has traded around $50 for months. If it jumps to $80 without anything changing at the company, the chance of a pullback toward the average rises. The price is not wrong because a formula says so, but because there are fewer buyers willing to keep adding at that level.

The idea comes from statistics. In the 1880s, Sir Francis Galton described it in his work on body height: children of very tall parents were, on average, closer to the average than their parents. Traders apply the same principle to market overreactions.

The logic is simple. Markets overshoot: euphoria pushes prices too high, panic pushes them too low, and neither extreme lasts long. As a trader, you look for markets whose price is far from its average and take a position against the current move. Our guide to range trading shows the same idea inside a sideways market.

The market phase decides. Mean reversion works best in sideways markets. Strong trends keep shifting the average up or down, so a stock in an uptrend keeps setting new, higher averages. That is where the classic logic fails, and it fails systematically, not only now and then.

The return also does not come on schedule. Sometimes it takes days, sometimes weeks, and if you enter too early you sit through the move or get stopped out. A short-term example with fixed rules is the Turtle Soup method, which trades failed 20-day breakouts.

The math behind it, for advanced readers

Mean reversion can be measured, which makes the search for setups systematic instead of subjective. The average is usually a moving average: a 20-day simple average adds the last 20 closes and divides by 20, and every current price is compared with that reference.

To compare deviations across markets, quantitative traders use the z-score. The z-score says how many standard deviations a price is away from its average. A value above 2 or below minus 2 counts as an extreme deviation. Its advantage is that a z-score of 2 means the same in gold as in EUR/USD, even though their price levels are completely different.

How fast a price returns is described by the half-life. In the Ornstein-Uhlenbeck model, a speed parameter called theta measures how strongly price is pulled back; in this continuous-time model with a positive theta, the half-life, the natural logarithm of 2 divided by theta, is the time in which the expected deviation shrinks by half. It is a property of the model and the data it was estimated on, not a promise about any single move; estimated on daily data, it says little about intraday behavior.

Whether a series tends to return at all can be tested. The Augmented Dickey-Fuller test checks, under its statistical assumptions, whether the null hypothesis of a unit root can be rejected, which would point to a stationary series that fluctuates around a stable average. A rejection does not mean that trading the series is profitable. For everyday trading, the basic sequence is enough: calculate the average, measure the deviation, trade the extremes.

How does mean reversion trading work?

A mean reversion trade follows four steps that build on each other. You define the average, measure the deviation, enter at an extreme and exit near the mean.

  • Define the average: a simple or exponential moving average gives the reference line, for example the 20-day average.
  • Measure the deviation: Bollinger Bands draw two standard deviations around the average; the RSI measures the move on a scale of 0 to 100.
  • Enter at an extreme: far below the average, the market counts as oversold, and the chance of an upward correction rises; above it, the reverse applies.
  • Exit near the mean: conservative traders close at the average; aggressive versions wait for the full countermove and accept more risk.

Many typical mistakes come from entering during a running trend. Three conditions make it more likely that the approach can work at all:

  • Stationarity: the market must fluctuate around a stable average. In trends, the average keeps moving, and a “cheap” price often gets cheaper.
  • Liquidity: enough volume and tight spreads matter, because costs in thin markets eat the small profit before the trade works.
  • Sideways phases: in a range, the average stays stable and outliers correct more reliably.

When does mean reversion work best?

The time frame changes the hit rate noticeably. Intraday charts from 5 minutes to 1 hour give several setups a day, especially in forex and futures, at the price of constant watching. Daily charts filter more noise; positions typically run three to ten days, which suits a full-time job.

Weekly charts give the fewest but calmest signals. Holding periods run from several weeks to months. As a rule of thumb, the time frame should fit how fast the deviation you trade tends to shrink. A half-life estimated on daily data describes daily swings; whether the same market reverts on a 5-minute chart has to be checked on intraday data.

Not every market suits the approach equally. Major currency pairs such as EUR/USD often move sideways, because high liquidity and opposing fundamental forces limit the swings. Commodities such as oil or gold also show reversion tendencies, above all in phases without a clear macro driver.

In stocks, selection decides. Large, liquid stocks are cheaper to trade than small caps, because spreads are tighter, and defensive sectors often swing more steadily than volatile tech names. Large indices such as the S&P 500 rise over the long run, so their price is not a stationary series; what traders use there are short-term pullbacks inside the trend, which is exactly what our test below does.

Mean reversion vs. trend following

Mean reversion and trend following are opposite philosophies. One bets on reversal, the other on continuation. Mean reversion buys weakness and sells strength; trend following buys strength because trends tend to continue.

The risk profiles differ just as much. Mean reversion produces many small wins with a high hit rate, and now and then a large loss when the correction does not come. Trend following produces many small losses and a few large winners, and the two feel very different to trade.

The practical answer is often both. Nobody knows in advance which market phase comes next, so many experienced traders combine the two approaches; our guide to momentum trading covers the trend side. Toby Crabel’s fund is one disclosed example: as of June 2022, its program statistics split the style into 55 percent momentum, 21 percent counter trend and 13 percent trend following, according to our portrait of Toby Crabel.

The ADX helps to classify the phase. Readings below 25 suggest a sideways market, readings above it a trend. The ADX lags the move, though, so market structure remains the more important signal.

Five indicators for mean reversion

The right indicators separate a rule-based approach from guessing. These five are the most widely used, and each one draws a line or a scale around the average. Our overview of trading strategies shows where mean reversion fits among other approaches.

  • Moving averages: the simple average weights all prices equally; the exponential average weights recent prices more and reacts faster. An average alone gives no threshold for an extreme.
  • Bollinger Bands: a 20-day simple average with bands two standard deviations above and below it. A touch of the outer band shows a stretched market, but in strong trends price can ride the band for days.
  • RSI: the Relative Strength Index measures recent gains against losses on a scale of 0 to 100; readings above 70 count as overbought, below 30 as oversold. Shorter settings such as RSI(2) or RSI(4) give more signals.
  • Williams %R: developed by Larry Williams, it shows where the close sits in the recent range on a scale of 0 to minus 100, the same quantity as the stochastic oscillator turned upside down. Readings above minus 20 count as overbought, below minus 80 as oversold.
  • Keltner Channels: bands a multiple of the Average True Range above and below an exponential average. They run smoother than Bollinger Bands and react less to single outliers.

For intraday traders, the VWAP is a common reference line. It weights the average by volume, and many day traders treat it as the session’s mean; our VWAP guide shows how it is calculated.

Our test: two RSI rules on SPY since 1993

Many published mean reversion backtests show hit rates of 80 percent or more, often from a few dozen trades. We wanted to see what two common rule sets did over a long period, so we tested them ourselves on SPY, the S&P 500 ETF. The rules follow the short-term RSI pullback strategies popularized by Larry Connors; we fixed the exact thresholds before looking at any result.

Rule Entry Exit
A Close above the 200-day average and RSI(10) below 30; buy at the next open Next open after RSI(10) rises above 40, or after 10 trading days
B Close above the 200-day average and RSI(4) below 25; buy at the next open Next open after RSI(4) rises above 55

Daily SPY data from Yahoo Finance, dividend-adjusted, signals from November 12, 1993 (the 200-day average is first available on November 11) to November 20, 2026. RSI with Wilder smoothing. Long only, no stop, each trade reinvests the full capital, idle cash earns nothing, and a fixed 0.05 percent cost per side stands in for commissions and spread; there is no separate slippage model. Own calculation.

Bar chart of our own test on SPY daily data from November 1993 to November 2026. Return a year compounded, Rule A 1.7 percent, Rule B 5.1 percent, buy and hold 10.4 percent. Share of days in the market, Rule A 2.3 percent, Rule B 12.6 percent, buy and hold 100 percent. Rule A, RSI(10) below 30, won 80 percent of 49 trades; Rule B, RSI(4) below 25, won 77.5 percent of 204 trades. Both buy at the next open and only above the 200-day average, no stop, 0.05 percent cost per sideClick to enlarge
Both rules won most of their trades, but they were rarely in the market, so their total return stayed well below buy and hold. Own backtest on dividend-adjusted SPY daily data, November 12, 1993 to November 20, 2026; rules fixed before the run; past results, not a forecast.

Source: Kagels Trading, own backtest.

Both rules won most of their trades. Rule A won 80 percent of 49 trades, about one and a half signals a year; Rule B won 77.5 percent of 204 trades. The average trade made 1.14 and 0.82 percent. On those numbers, both look like the high hit rates that make mean reversion popular.

The total return tells a different story. Rule A compounded to 1.7 percent a year and Rule B to 5.1 percent, against 10.4 percent a year for simply holding SPY. The main reason is exposure: the rules were in the market on only 2.3 and 12.6 percent of all days, and in our calculation the cash between trades earned nothing. Per day invested, they did well; as a whole strategy, they left most of the market’s return on the table.

The losses also show the typical profile. In Rule B, the average win was 1.54 percent, but the average loss 1.64 percent, and the worst trade, from July 28 to August 16, 2011, lost 8.6 percent. With no stop, a single correction that keeps going can cost several winners.

Read these numbers as a sample, not a promise. They come from one market that rose strongly over the period, with rules that only buy above the 200-day average. In a sideways or falling market, the picture would look different. Our guide to backtesting explains why a small number of trades and a single market say little about the future.

A chart-based alternative: the 1-2-3 reversal

Not every mean reversion approach needs an indicator. The 1-2-3 pattern, which Victor Sperandeo described in his book “Trader Vic: Methods of a Wall Street Master”, defines a trend reversal by three points on the chart.

  • Point 1: the last extreme of the trend, the lowest low in a downtrend.
  • Point 2: the first countermove that breaks the trendline.
  • Point 3: the failed attempt to continue the old trend; the closer it is to point 1, the stronger the signal.
  • Entry and stop: enter on the break of point 2, with the stop just beyond point 3.

Unlike the two RSI rules, this pattern comes with a price-based stop. The stop defines the intended exit and the planned risk of each trade, which suits traders who do not want an open-ended risk; gaps and slippage can still make the actual loss larger.

Pairs trading: mean reversion between two markets

Pairs trading is a mean reversion strategy that trades the price gap between two closely related securities instead of a single price. Two stocks from the same industry move largely in parallel for years. If one suddenly runs ahead while the other lags, the gap widens beyond its usual range. You buy the laggard and sell the leader short; if the gap closes, the pair makes a net profit, even if one of the two positions loses.

The aim is to reduce market exposure. One position is long and one is short, but that alone does not remove the market risk; how neutral the pair really is depends on the position sizes and the hedge ratio between the two. A high correlation also does not prove that the spread is stationary. The gap is usually measured with the z-score of the spread: above 2 signals an unusually wide gap, and the position is closed as it moves back toward 0. The statistical groundwork is larger than with any other approach here.

The main risk is a broken relationship. After a takeover or a profit warning, the historical link can break for good, and the spread keeps widening while both sides lose. A stop on the spread is therefore a must, even though many single-market mean reversion rules do without a stop; our guide to the stop loss order covers the order types.

What does the research say?

The academic base is older than most trading blogs. In 1988, James Poterba and Lawrence Summers reported in “Mean reversion in stock prices”, published in the Journal of Financial Economics, evidence that stock returns are negatively autocorrelated over longer horizons: strong phases tended to be followed by weaker ones. They also noted that a random walk could not be rejected at conventional significance levels, so the evidence is suggestive rather than conclusive.

In the same year, Eugene Fama and Kenneth French reported a similar structure. In “Permanent and Temporary Components of Stock Prices” in the Journal of Political Economy, they split prices into a permanent and a temporary component; only the temporary part reverts to the mean. Both studies work with returns over several years, not with the short-term swings that most traders trade.

The evidence is mixed, and the execution decides. Profitability depends on three points: the right market, disciplined execution and consistent position sizing. With a hit rate of 70 percent and losses larger than wins, a handful of big losses can ruin a year, so position size matters more than the entry signal.

Pros and cons of mean reversion trading

Mean reversion has clear strengths and structural weaknesses. You should know both before you put real money on the line, because the weaknesses show up rarely but hard.

Pros of mean reversion

  • High hit rate: most trades end in profit, which makes the approach easier to stick with.
  • Clear rules: entries and exits can be expressed in numbers, which reduces gut decisions.
  • Easy to automate: the rule-based logic fits backtests and trading systems.
  • Research base: two independent 1988 studies found mean reversion in stock prices.

Cons of mean reversion

  • Small wins per trade: losses are often larger than wins, so single losses weigh heavily.
  • Fails in trends: in directional markets, losses pile up because the average moves along.
  • Stops are awkward: tight stops cut off the countermove, wide stops raise the maximum loss.
  • Low exposure: in our test, the rules were rarely in the market and lagged buy and hold.

Who is mean reversion for?

Mean reversion suits traders who already follow trends. The two approaches earn in different market phases, which can stabilize the overall result. It works best for swing and position traders on daily and weekly charts; very short-term traders pay far more in costs.

It does not suit everyone. If you cannot buy into a falling market without panicking, you will break the rules when it counts. And if you trade without a fixed risk limit, the first real trend phase can cause serious damage; keep every trade in a trading journal so you see that early.

Conclusion: is mean reversion the right strategy for you?

Mean reversion works, but not everywhere and not all the time. The effect is documented in research, and in our test both rules won four out of five trades. At the same time, both earned far less than simply holding the market, because they were rarely invested.

Bjarne’s advice from systematic trading: treat mean reversion as one part of a system, not the whole system. Pick one rule, test it on your market with enough trades, and check the exposure and the size of the losses, not only the hit rate. Start small, document every trade and increase risk only after several months of measured results.

Frequently asked questions about mean reversion trading

What does mean reversion mean?

Mean reversion means that prices tend to move back toward their average after an extreme move. If a stock rises from $50 to $80 without any fundamental change, the chance of a correction toward its historical average rises.

Which indicators are best for mean reversion?

The most common are Bollinger Bands, the RSI, Williams %R, Keltner Channels and moving averages. They mark how far price has moved from its average. Quantitative traders also use the z-score to compare deviations across markets.

Does mean reversion trading work?

It can, in the right market phase. In our test on SPY from 1993 to 2026, two RSI rules won 78 to 80 percent of their trades, but returned 1.7 and 5.1 percent a year against 10.4 percent for buy and hold, because they were rarely in the market.

What is the difference between mean reversion and trend following?

Mean reversion bets on a reversal; trend following bets on continuation. Mean reversion has a high hit rate with small wins and occasional large losses; trend following has many small losses and a few large winners.

What time frame is best for mean reversion?

It depends on how fast the market returns to its average. Daily charts with holding periods of three to ten days suit most traders. The time frame should fit how fast the deviation you trade tends to shrink, which you can estimate as a half-life on data of that time frame, the time in which the expected deviation halves.

Is pairs trading a form of mean reversion?

Yes. Pairs trading bets that the price gap between two related securities returns to its usual range. It aims to reduce market exposure, but its neutrality depends on position sizes, and it carries the risk that the relationship between the two breaks for good.

This English edition is based on our German edition on kagels-trading.de and has been adapted for international readers.

← All articles