Mean Reversion Strategy: How to Trade the Snap Back

What mean reversion means in practice

A mean reversion strategy buys after an unusually sharp decline and sells after an unusually sharp advance, expecting price to return toward a recent average. The reference "mean" is short term: a 5 or 10 day moving average, a Bollinger band midline, or the prior week’s range midpoint. It is not a claim about intrinsic value.

The holding period is short, usually one to ten days, because the imbalance being exploited is temporary. The longer you hold, the more you are exposed to the possibility that the move was information rather than noise, which is precisely the risk that makes this family dangerous.

Why short-term reversal exists

• Liquidity demand. A fund forced to sell a large block quickly pushes price below where it would otherwise settle. Buyers who supply that liquidity are compensated when the pressure ends. This is the core mechanism.
• Overreaction to news. Initial reactions to headlines frequently overshoot, particularly when amplified by stop cascades and margin liquidations.
• Market maker inventory. Dealers who absorb one-sided flow must offload inventory, and they widen quotes to attract the other side, creating a short-lived price concession.
• Option hedging flows. Dealer gamma hedging can mechanically buy dips and sell rallies in index products, dampening moves within a range.
• Index and rebalancing effects. Scheduled flows create temporary distortions that reverse once the mechanical buying or selling finishes.

Every one of these is a genuine service being paid for. Notice also that each disappears in exactly one circumstance: when the selling is informed rather than mechanical. That is why mean reversion is profitable most of the time and occasionally catastrophic.

A complete rule set for index mean reversion

Universe

Liquid broad index ETFs, or large-cap stocks with average daily dollar volume above 50 million USD and no earnings within 3 sessions.

Timeframe

Daily bars, evaluated on the close, executed at the close or next open.

Regime filter

Only take long entries when the close is above the 200-day simple moving average. This single rule removes most of the worst losses.

Entry trigger

RSI(2) below 10, or price closing below the lower Bollinger band (20, 2.0), or a decline of at least 2 ATR(10) from the 5-day high. Use one, not all three.

Entry execution

Buy at the close of the signal day, or at a limit below the close for a better price at the cost of missed trades.

Profit exit

Sell when the close crosses back above the 5-day moving average, or when RSI(2) exceeds 70.

Time stop

Exit unconditionally after 8 trading days. A reversion that has not occurred was not a reversion.

Disaster stop

Hard stop at 3 ATR(10) below entry. It will rarely trigger and it is the reason the account survives the trade that does not revert.

Position size

Risk 0.5 percent of equity to the disaster stop. Maximum 4 concurrent positions, maximum 1 per sector.

Worked example with the arithmetic

Equity 50,000 USD, risk 0.5 percent, so 250 USD per trade. A broad index ETF trades at 512.00, above its 200-day average at 486. It falls four consecutive days and closes at 496.50 with RSI(2) at 4. ATR(10) is 6.20.

Note the payoff shape. The win was 0.6R, not 3R. Mean reversion accumulates modest gains, and its profitability depends on the frequency of those gains and on the rarity of the disaster stop being hit. A single 3R loss requires five average winners to repair, which is why position size and the trend filter carry all the weight.

The risk profile, stated plainly

That table deliberately shows a losing configuration, because it is what happens when the disaster stop is hit 10 percent of the time instead of 3 or 4 percent. The entire viability of the strategy rests on how often the extreme case occurs, which is determined by the regime filter, the instrument choice, and whether you trade through earnings and known events. A version that looks excellent on liquid index ETFs can be negative on single stocks with identical rules.

Variations and where each fits

How to test it without fooling yourself

• Include the crises. Any test that excludes 2008, March 2020, and 2022 tells you nothing about the risk you are taking. The strategy looks superb between crises by construction.
• Include delisted securities. Testing mean reversion on today’s index constituents guarantees you never buy a company that went to zero. This is survivorship bias and it is devastating here.
• Model the fill honestly. Buying at the close of a sharply down day means buying into the closing auction at a price you may not get. Test with next-open execution as a conservative alternative.
• Check earnings exposure. A large share of single-stock disaster stops occur on earnings gaps. Exclude the window and re-measure.
• Report the maximum single-trade loss, not only the average. The tail is the strategy’s real characteristic.

Mistakes that turn a good strategy into a blow-up

1. Trading without a stop because "it always comes back". It comes back until the one time it does not, and that time removes the previous two years of gains.
2. Averaging down freely. Adding at each new low without a capped total risk converts a 0.5 percent position into an account-defining one.
3. Removing the trend filter after a good year. The filter costs return in bull markets and saves the account in bear markets. Judging it during a bull market guarantees the wrong conclusion.
4. Trading single stocks the same way as indices. An index reverts because it is a portfolio; a single company can have a permanent problem.
5. Increasing size after a long winning streak. High win rates create false confidence precisely before the environment changes.
6. Holding past the time stop. The time stop is what converts an unbounded thesis into a bounded trade.