What it is
Mean reversion to the 20-day moving average is the swing trader's counterpart to the trend pullback: instead of buying dips in a trend, it buys extremes in a range and sells them back at the average. The stretch is defined by a volatility band, either bollinger-bands at 2 standard deviations or a 2 ATR envelope. The trade lasts 2 to 8 days and the target is the 20-day average itself.
This setup has a high win rate and small wins, with occasional large losses when a range turns into a trend. It is honest to say that the risk profile is the opposite of what most traders want.
The logic
In a range, price oscillates because no participant has enough urgency to move it out. When a short-term move stretches to the edge of the band, it has usually been driven by short-term traders chasing a small piece of news or a technical break. Those traders have weak conviction, and when the move stalls, their exits pull price back toward the average, where the longer-horizon participants are still content to trade. The average is the target because it is the level the range's real participants keep returning to.
On the other side are breakout traders and momentum chasers who bought the stretch. When the range holds, they are wrong and their exits are your profit. When the range breaks, they are right, and you are the one paying.
Setup rules
- Market: liquid index ETFs, sector ETFs, large-cap stocks with low idiosyncratic news; major forex pairs. Avoid small caps and anything with a pending catalyst.
- Timeframe: daily.
- Range conditions: the 20-day MA has a slope near zero (less than 2 percent change over 20 days); price has crossed the 20-day MA at least 3 times in the last 40 days; the 50-day and 200-day averages are within 5 percent of each other. A flat ADX-style trend reading helps but is not required.
- Stretch conditions: a daily close below the lower 2-standard-deviation band (for longs), at least 2 ATR below the 20-day MA; rsi (2-period) below 10 as a secondary check.
- Disqualifiers: the stretch was caused by a gap on news; earnings within the holding period; the index itself is in a strong trend.
Entry, stop, target
Enter on the next day's open after the close below the band, or on a limit order at the band if you want a better price at the risk of missing the trade. Stop is 1.5 ATR below the entry; because reversion trades often go against you before working, a tighter stop kills the win rate. Target is a close at or above the 20-day MA. Time stop: exit after 8 days regardless.
| Item | Level | Notes |
|---|---|---|
| 20-day MA | 100.00 | Flat |
| Lower band | 95.00 | 2 SD, ATR is 2.00 |
| Entry | 94.60 | Next open after close below band |
| Stop | 91.60 | 1.5 ATR, risk 3.00 |
| Target | 100.00 | 20-day MA, reward 5.40, 1.8R |
Do not scale out; the trade is small enough that one exit is fine. Do not hold past the average hoping for the upper band; that converts a reversion trade into a trend trade with no thesis.
Position sizing and risk
Risk 0.5 percent of equity per trade at most, sized at /tools/position-size. Mean reversion positions are correlated across the market during a selloff: if five ETFs close below their lower bands on the same day, that is one trade, not five, and it should be sized as one. Set a portfolio cap on simultaneous reversion positions and respect the portfolio-heat limits in /learn/risk-management.
What breaks it
- Trend emergence. The range that breaks is the whole risk. The first two stretches after a range turns into a trend are exactly the ones this playbook buys, and they are the ones that lose 1.5 ATR each. A long-term equity curve for mean reversion is a staircase up with a cliff every year or two.
- Tail risk. Because the stop is wide and the position is held overnight, a gap of 3 to 5 ATR on a shock day produces a loss several times larger than planned; see tail-risk.
- Costs. Low per trade, but the win is small (1.8R on a good day, often less), so the spread and commission are a larger share of profit than for trend trades.
- Edge decay. Short-term mean reversion in US equity indices was one of the strongest documented effects of the 2000s and has weakened noticeably since. It is still present but smaller; assume any backtest overstates it.
How to test it
Test on at least 20 years of daily data across a basket of ETFs and large caps with realistic costs and no look-ahead on the band calculation. Report the win rate, average win, average loss and profit-factor, and pay particular attention to the worst 5 trades and the worst 3 months; those dominate the real outcome. Segment by whether the 20-day MA slope was flat, since that is the key filter. Minimum 400 trades. Then paper-trade for at least 6 months, because the setup can look great in a calm year and lose it all in a volatile quarter.
Variations
- RSI(2) system: a well-known simple version using only a 2-period RSI reading below 5 and a close above the 200-day for direction.
- Intraday version: 5-minute VWAP bands; see vwap-reclaim-reject.
- Pairs version: revert the spread between two related instruments instead of a single price; see pairs-spread-trading.
Further reading
mean-reversion, bollinger-bands, moving-average, rsi, range, atr, tail-risk, profit-factor, chop, buy-the-dip.
Related playbooks: ema-pullback-trend, three-day-pullback, range-day-playbook, funding-rate-mean-reversion