TL;DR
Past losers tend to rebound — but reversal is two different phenomena at two horizons. Short-term (1-week to 1-month) reversal is largely liquidity provision / microstructure and is high-turnover. Long-term (3–5 year) reversal is the overreaction story that helped found behavioral finance — and largely overlaps the value premium. Conflating them is a common error.
A 40-year arc
De Bondt & Thaler (1985) — long-term reversal: extreme 3–5-year losers outperform winners over the next 3–5 years; evidence of overreaction.
Jegadeesh (1990); Lehmann (1990) — short-term reversal: last month's losers beat last month's winners next month.
Jegadeesh & Titman (1993, 2001) — momentum sits between the two reversals (3–12 months), and only a partial long-horizon reversal appears at years 4–5.
Microstructure work — short-term reversal is tied to bid-ask bounce, inventory, and the price of immediacy, not mispricing per se.
Sub-threads
Short-term (1-week / 1-month) reversal · long-term (3–5 year) reversal · the momentum "sandwich" between them · industry/sector reversal.
Why it works
Short-term — compensation for providing liquidity: market makers and contrarians earn the spread/immediacy premium as transient price pressure reverses.
Long-term — overreaction: investors extrapolate good/bad news too far, then prices mean-revert (De Bondt-Thaler); this is largely the value effect at long horizons.
The dark side
Short-term reversal is costly — its gross Sharpe is high but turnover is enormous; net of transaction costs much of it is really the liquidity-provision return, not free alpha.
Long-term reversal ≈ value — little marginal alpha after controlling for book-to-market.
Microstructure noise — short-horizon results are sensitive to bid-ask bounce and stale prices.
Does it survive out of sample?
Short-term reversal persists gross but is marginal net of costs (it's a liquidity-provision strategy); long-term reversal persists but is largely subsumed by value. Our replications recompute both short-term and long-term reversal and score them on the holdout.
Run it yourself
Replications — Jegadeesh short-term reversal, De Bondt-Thaler long-term reversal.
Curriculum — the synthetic market ships a reversal_1w factor; Mission 9 (pairs trading) covers the mean-reversion machinery; the cost-of-trading mission shows why short-term reversal is so cost-sensitive.
Playground / Competitions — build a reversal signal and score it net of costs.