🎯 In 30 seconds: Mean reversion is the bet that a price that moved unusually far from its normal level will move back toward it. Traders buy short-term dips in strong stocks, sell near the top of ranges, or trade two related stocks against each other. It tends to win often but in small amounts, and the losses come when a "dip" turns out to be the start of a real decline. Mean reversion works best over days to a few weeks; momentum works better over months. Mixing up the two time scales is a common way to lose.
The idea in plain words
The mean is an average. In trading it usually means a moving average or a typical relationship. Mean reversion means a price that has stretched far from that average tends to drift back toward it.
Why would that happen? Short-term moves are often driven by crowds overreacting, a big fund selling in a hurry, or temporary fear. When the pressure passes, the price can recover part of the move. Lehmann's research on US stocks found that the previous week's biggest winners and losers tended to reverse the following week.
A mean reversion trader is the person buying when others are selling in a hurry. That is uncomfortable by design.
Three common setups
1. Oversold bounce in an uptrend. The stock is above its rising 200-day moving average, so the long-term trend is up. It then falls sharply for a few days, for example closing below its lower Bollinger Band (see Volatility tools) or with a very low short-term RSI (see Momentum indicators). You buy and expect a bounce back toward the 20-day average within days.
2. Range trading. The stock has bounced between clear support and resistance for months. You buy near support and sell near resistance. See Support and resistance.
3. Pairs trading. Two stocks that normally move together drift apart. You buy the one that fell behind and short the one that ran ahead, expecting the gap to close. Covered in detail below.
Worked example: an oversold bounce
Hypothetical. Company X trades above its rising 200-day average. Its 20-day average is $100.00 and the standard deviation of the last 20 closes is $2.50.
| Lower Bollinger Band: $100.00 minus 2 x $2.50 | $95.00 |
| After three down days, the stock closes at | $95.00 |
| Entry at the close | $95.00 |
| Target: back to the 20-day average | $100.00 (+$5.00) |
| Stop: a fixed amount below, decided in advance | $91.00 (-$4.00) |
| Time stop: exit after 5 trading days regardless | day 5 close |
With a $15,000 account and 1% risk ($150): shares = $150 / $4.00 = 37. Position = 37 x $95 = $3,515.
Notice the reward to risk here is only 1.25 to 1. Mean reversion systems often have reward smaller than or close to risk and need a high win rate, often well above 50%, to make money. That is the reverse of trend following. See Risk/reward and expectancy for the breakeven win rates.
The time stop matters. If the bounce does not come within a few days, the reason for the trade is gone, even if the price stop has not been hit.
Pairs trading in plain words
Pick two companies that do almost the same thing and are hit by the same news, such as two large firms in the same industry. Their prices tend to move together, so the ratio of their prices stays in a fairly stable range.
When the ratio moves unusually far from normal, you short the expensive one and buy the cheap one in equal dollar amounts. If the whole market rises or falls, the two legs mostly cancel out. You profit only if the gap between them closes.
To measure "unusually far", traders use a z-score:
z = (today's ratio minus average ratio) / standard deviation of the ratio
Worked example (hypothetical). Over the past year, the price ratio of Stock P to Stock Q has averaged 2.00 with a standard deviation of 0.05.
| Today: P = $106.00, Q = $50.00, ratio | 2.12 |
| z-score: (2.12 minus 2.00) / 0.05 | +2.4 |
| Rule: enter when z is beyond 2, exit near 0, stop beyond 3.5 | Enter |
| Short P: 47 shares x $106.00 | $4,982 |
| Buy Q: 100 shares x $50.00 | $5,000 |
Three weeks later P = $104.00 and Q = $52.00. The ratio is back to 2.00 (z = 0), so you close both legs.
| Short P: 47 x ($106.00 minus $104.00) | +$94 |
| Long Q: 100 x ($52.00 minus $50.00) | +$200 |
| Total before costs | +$294 |
The stop at z = 3.5 is a ratio of 2.175. If the gap keeps widening to there, you close both legs and accept the loss, because the relationship may have changed.
Research supports the basic idea, with caveats. Gatev, Goetzmann and Rouwenhorst tested a pairs rule on US stocks from 1962 to 1997 and found average annualized excess returns of up to about 12% for the top pairs. They also noted some of the profit may come from market microstructure effects such as bid-ask bounce. Results from older data do not guarantee the effect still exists after costs today.
Why mean reversion and momentum both exist
It sounds like a contradiction: momentum says winners keep winning, mean reversion says they pull back. The answer is time scale.
| Horizon | Pattern found in research | Style that fits |
|---|---|---|
| About 1 week to 1 month | Short-term reversal: recent big movers tend to partly reverse | Mean reversion |
| About 3 to 12 months | Momentum: winners tend to keep beating losers | Momentum, trend following |
| Several years | Long-term partial reversal | Value investing |
A practical rule that many traders use: only take mean reversion trades in the direction of the longer trend. Buy short dips in stocks that are in long-term uptrends. Do not buy dips in stocks in long-term downtrends, which is where the next section's problem lives.
The falling knife
"Catching a falling knife" means buying a stock that is dropping fast because it looks cheap, and then watching it keep dropping. Oversold can become more oversold for weeks.
How to avoid it:
- Trend filter. Only buy dips above a rising 200-day average.
- Hard stop and time stop. Decide both before entry.
- No averaging down on a losing mean reversion trade. Adding to a loser turns a small planned loss into a big unplanned one (Beginner mistakes).
- Check why it fell. A drop on fraud allegations, a lost lawsuit or a guidance cut is a change in the business, not a temporary overreaction.
Try it
In the backtest tool, test a simple dip-buying rule on a broad index ETF: buy after the price closes below its lower Bollinger Band while above the 200-day average, and sell after 5 days. Note the win rate, the average win, the average loss and the worst single trade. Then remove the 200-day filter and run it again. Does the worst trade get worse? Record both runs in your trading journal.
- Over what time scales do mean reversion and momentum tend to show up? Answer: Short-term reversal shows up over about 1 week to 1 month. Momentum shows up over about 3 to 12 months.
- A stock’s 20-day average is $100.00 and the standard deviation of the last 20 closes is $2.50. Where is the lower Bollinger Band? Answer: $95.00 ($100.00 minus 2 x $2.50).
- The P to Q price ratio has averaged 2.00 with a standard deviation of 0.05. Today it is 2.12. What is the z-score, and does the example rule say enter? Answer: +2.4 ((2.12 - 2.00) / 0.05). Yes: the rule enters when z is beyond 2.
- How do you avoid catching a falling knife? Answer: Only buy dips above a rising 200-day average, set a hard stop and a time stop before entry, never average down on the loser, and check why the stock fell.