Mean Reversion

Mean reversion bets on a return toward a reference: a moving average, VWAP, the middle of a range, a band. The setup is a departure that you judge to be too far, plus a trigger that the departure is stalling. It is the conceptual opposite of trend following. In a stable range it harvests small, frequent wins. In a trend it sells the strength and buys the weakness and gets run over. The reference has to be defined. "Too far" without a number is a mood. A close two ATR from the average, a touch of the lower band at a level, a gap beyond the prior range: those are definitions. They can be tested. A feeling that the candle is long cannot.

Choose the reference and the stretch. Then choose the trigger that keeps you from buying a market that is still falling. A reversal bar, a reclaim of a minor level, a stop in the order book. Entering at the band with no trigger is volunteering to be early. The target is the mean, not a new trend. Take the profit at the mean. A mean-reversion trade that you hold for a trend is a winner you converted into a different bet, often giving the mean-reversion profit back. The stop is beyond the stretch, at a price where the "too far" thesis is simply wrong. If that stop is huge, the stretch is a shock, not a fade, and the size should be tiny or zero.

Run mean reversion in instruments and hours that actually oscillate: index products in the middle of a dull session, a range on the daily chart, a pair that has a real economic link. Turn it off on event days and after a volatility regime shift, when the mean itself is moving. Combine with a trend filter if you want fewer disasters. Only fade stretches that are against a flat higher timeframe, not against a waterfall. The best mean-reversion trades look almost boring. If the trade needs a heroic reversal, it is a reversal trade. Count it as one.

Fading a strong trend because a band was touched. Bands are touched by trends every week. Also moving the mean after the entry so you can claim the target was hit. And averaging down a mean-reversion loser. You are adding to a market that has not returned, which means the premise is failing, and you are making it larger.

The mean moves. A strategy anchored to yesterday's average loses when today's fair price is somewhere else, which is exactly what a trend is. Fat tails make the rare loss larger than the frequent gains. Without a hard stop and a daily cap, mean reversion is a method that looks profitable until the one day it is not. Backtests that fade every band touch and exit at the average, with no costs and no "still falling" filter, overstate what you can take home.