Skip to main content

Swing Setups: Mean-Reversion

Fading short-term extremes back toward value.

Updated Jun 23, 2026 at 8:47pm

  • 1826e57ddeff Oversold Bounce (RSI-2) 1 771
  • 1257e3eb3230 Bollinger-Band Reversion 1 828
  • 18250c58eb22 Gap Fill 1 905
  • 1827076d8bd4 Snap-Back to VWAP / Moving Average 1 709
  • 1828ce872471 Dead-Cat Bounce (Short) 1 776
Tree Key
Expandable — has sub-topics
475Local Id for node
a1b2c3d4Click to see full UUID
5Sub-topics
6Documents
5.2k wordsResearch depth
5Open node
Research Draft Medium 1,166 words

Mean-reversion setups bet on the opposite force from breakouts and continuation trades: that a short-term price extreme will snap back toward a recent average — a moving average, VWAP, a Bollinger midline, or a prior price level — rather than keep extending. The trader enters into weakness (or strength) at the extreme, anticipating a reflexive bounce, and exits at "the mean." Holding periods for swing-grade mean reversion run from one to a handful of days. The defining tension of the whole family: it is statistically supported in range-bound conditions and actively dangerous in strong trends, where "oversold" simply gets more oversold.

The setups

  • Oversold bounce (RSI-2 / Connors). The best-known codified mean-reversion swing setup. Larry Connors' 2-period RSI (StockCharts ChartSchool) buys only when price is above its 200-day SMA (so the bigger trend is up) and RSI(2) drops to a deeply oversold level — Connors found dips below 5 outperformed dips below 10. It is therefore a pullback-within-an-uptrend tactic, not a bottom-picker for falling knives.
  • Bollinger-band reversion. Bollinger Bands (John Bollinger, 1980s) plot a 20-period SMA midline with ±2 standard-deviation bands. A close stretched to or beyond the lower band flags a statistical extreme; the reversion trade targets a snap-back to the midline. A common risk-reducer is waiting for price to close back inside the band before entering, confirming the reversal has begun (TrendSpider, Investopedia).
  • Gap fill. An opening gap leaves an "empty" price zone that price often retraces to close. Fill probability is conditional, not guaranteed: small, low-volume "common" gaps fill far more often than catalyst-driven breakaway gaps, and choppy markets fill gaps more readily than strong trends. (See Adoption, debate & evidence for sourced numbers.)
  • Snap-back to VWAP / moving average. Intraday and short-swing traders treat VWAP (volume-weighted average price, reset daily) or a 20–50 SMA/EMA as the "fair value" magnet: buy a stretch below, sell a stretch above, exit at the average (TradingSim). VWAP is the institution-favored intraday mean because it embeds volume.
  • Dead-cat bounce (short). The aggressive, inverse case: shorting a bounce inside an established downtrend, expecting price to roll over and make new lows (Wikipedia, Britannica Money). This is the highest-risk member of the family — see limitations.

How it's used in practice

The operating recipe is uniform across setups: define the mean, measure the deviation, wait for the snap-back signal, exit into the mean. Entry is at the extreme into a support (or, for shorts, into resistance) — for example, RSI(2) below 5 at a rising 200-day SMA, or a tag of the lower Bollinger band at a prior swing low. The stop sits just beyond the extreme (below the support being faded). The target is the average itself — the 5-day SMA in Connors' rules, the Bollinger midline, VWAP, or the gap-origin price. Crucially, mean reversion is a limited-target game: you exit at the mean, you do not let it run, because the edge decays once price returns to fair value. Connors notably found that fixed stop-losses reduced returns in backtests because they ejected trades before the bounce arrived — a real finding, but one that trades drawdown control for win rate, and should not be copied blindly onto leveraged or illiquid names.

Adoption, debate & evidence

This family deserves an honest contrast with oscillator-based trend prediction (e.g. using RSI to call a new uptrend), which is largely folklore. Short-term reversal has comparatively stronger documented empirical support. The academic "short-term reversal" anomaly is well established: Jegadeesh (1990) reported that a strategy buying prior-month losers and selling prior-month winners earned roughly 2% per month over 1934–1987, and Lehmann (1990) found sizable weekly winner/loser reversals — both classic results (Jegadeesh & Titman context, Univ. of Houston PDF; summarized in Springer / Short-Term Reversal). On the practitioner side, Connors and Alvarez's published backtests popularized RSI-2 as a durable mean-reversion edge.

Three caveats are essential. (1) The edge may be liquidity, not psychology. Jegadeesh & Titman and later work argue much of the reversal premium reflects bid-ask bounce, short-term price pressure, and the cost of providing liquidity — meaning paper returns can evaporate after realistic transaction costs (NY Fed staff report). (2) Gap-fill numbers are environment-dependent and often loosely sourced. Practitioner studies commonly cite that the S&P 500 fills its daily gap on the order of ~60% of the time, with small common gaps filling more often and earnings-driven breakaway gaps filling much less often, but exact figures vary by study, sample window, and gap definition — treat any single percentage skeptically. (3) It is regime-dependent. Every source agrees: Bollinger and RSI-2 reversion "underperforms in trending markets" where price rides a band for weeks.

Strengths & limitations

Strengths. Among technical setups, mean reversion has the most credible academic and backtested support; rules are objective and mechanizable; win rates are typically high (many small wins); and it profits from the choppy, directionless conditions where breakout strategies bleed.

Limitations. It is destroyed in strong trends — the payoff profile is many small wins punctuated by occasional large losses when a "dip" becomes a crash, so a removed or loosened stop (as Connors' tests favored) creates real tail risk. Shorting a dead-cat bounce compounds this: short selling carries theoretically unlimited loss, the bounce is "almost impossible to identify in real time" and only confirmable in hindsight, and a crowded short can trigger a squeeze that is the bounce. The Connors RSI-2 itself is "early" — moves often continue past the signal — so StockCharts suggests waiting for RSI(2) to cross back above 50 before entering. And reported edges shrink after costs and may already be arbitraged in liquid large-caps.

Sources