Mean reversion is a bet that stretched prices come back. The market agrees with you right up until it doesn't, so the screen is only the first filter — what separates a usable candidate list from a list of ways to bleed out slowly is everything you check next.
What the screen should measure
- ▸Distance from a reference, not a guess. Z-scores against a rolling mean, Bollinger distance, or percentile rank within the symbol's own history — a number with a distribution behind it, not “looks oversold.”
- ▸Liquidity before entry. Average volume and spread determine whether the reversion you measured is one you can trade. A signal you cannot enter or exit cleanly is a statistic, not a strategy.
- ▸Volatility state. Reversion behaves differently in high-vol and low-vol regimes. Filtering or sizing by the current volatility environment keeps the screen from mixing two incompatible markets.
- ▸Time at the edge. How long the symbol has been stretched matters: reversion from a two-day stretch and from a two-month drift are different trades with different holding periods.
Where screens lie
Two habits inflate reversion results. First, testing only on symbols that still exist — the ones that never reverted because they went bankrupt are missing from your sample. Second, ignoring costs: a strategy that harvests small spreads pays those costs on every cycle, and commission assumptions that were “close enough” in a spreadsheet decide the outcome.
There is also the regime problem. A reversion screen run during a calm year will look brilliant and then fail spectacularly in the first violent trend, because the premise — prices oscillate around a mean — stopped being true. The honest version of the screen reports how it performed across different volatility environments, not just its average.
That is why the useful workflow is a loop, not a funnel: screen a wide universe, backtest the survivors with costs, validate out-of-sample, and re-run the screen when the regime shifts. The screen finds candidates; the validation decides which of them deserve capital.