Skip to main content

RSI / Stochastics for Swings

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,192 words

RSI (Relative Strength Index, Welles Wilder, 1978) and the Stochastic Oscillator (George Lane, late 1950s) are the two most widely used bounded momentum oscillators. Both compress recent price action into a 0–100 scale and flag "overbought" and "oversold" extremes. For the swing trader holding two to ten days, they answer a narrow question: is the short-term move stretched enough that a pullback or bounce is statistically likely? Their core tension is that the same extreme reading means reversal in a range and strength in a trend — so the indicator is nearly useless without the context layer (trend, location, structure) sitting on top of it. Used as a standalone overbought/oversold trigger they are weak; used as a timing filter inside a defined setup they are genuinely useful.

How they're calculated

RSI = 100 − [100 / (1 + RS)], where RS = average gain / average loss over the lookback. Wilder's default is 14 periods; canonical thresholds are 70 (overbought) and 30 (oversold). Shorter periods (2–5) are far more sensitive and are the basis of mean-reversion systems; longer periods (20+) smooth signals for position work. Source: StockCharts ChartSchool – RSI.

Stochastics measures where the close sits within the recent high–low range: %K = (Close − Lowest Low) / (Highest High − Lowest Low) × 100, with %D = 3-period SMA of %K as the signal line. The common "Full" setting is 14, 3, 3 (lookback, %K smoothing, %D smoothing). Thresholds are 80 / 20. Fast = raw; Slow / Full add smoothing for fewer whipsaws. Source: StockCharts ChartSchool – Stochastic Oscillator.

The key structural difference: RSI is built from the magnitude of gains vs losses; Stochastics is built from the position of the close within its range. Stochastics is therefore more reactive and better at catching short, sharp swing reversals; RSI is steadier and better at gauging the persistence of momentum.

The setups (swing context)

  • Pullback timing in an uptrend (the highest-value use). Filter for trend first: price above a rising 50/200-day MA. Then buy weakness, not strength — wait for Stochastics %K to dip below 20 and turn up, or RSI to retrace toward 40 and curl higher. The oscillator isn't predicting a reversal; it's timing entry into a continuation. Constance Brown's "bull market range" (RSI roughly 40–90 in uptrends) is the framework here: 40 acts as support, not 30 (per ChartSchool).
  • RSI(2) mean reversion (Connors/Alvarez). Price above the 200-day MA; buy near the close (or next open) when RSI(2) drops below 5 (aggressive) or 10 (less aggressive); ChartSchool's primary exit is a move above the 5-day SMA (a higher-RSI cross-back, e.g. above 65–70, is a common variant). Holds are short, commonly cited around 2–4 days. Published in Short Term Trading Strategies That Work (Connors & Alvarez, 2008). Note Connors notoriously runs this without hard stops, which is what creates the occasional outsized loss.
  • Stochastics %K/%D crossover at an extreme. A bullish %K-crosses-%D while below 20 near a support level is a classic swing entry trigger; the cross above the line, not the oversold reading alone, is the signal.
  • Divergence near structure. Price makes a lower low but the oscillator makes a higher low (bullish divergence) at a tested support — a warning the down-swing is losing fuel. Most reliable at the end of a move into support/resistance, not in the middle of a trend.
  • Lane's bull/bear set-ups. A bull set-up = price lower high while Stochastics makes a higher high, often preceding a final pullback then a strong leg up (per ChartSchool). These are anticipatory, lower-probability, and should be confirmed by a break.

Failure modes a pro watches for: selling every overbought tag in a strong uptrend (the indicator can pin above 70/80 for weeks); fading divergences against a powerful trend (they fire repeatedly before any real reversal); and trusting an oversold reading with no support level or trend backing it — "oversold can get more oversold."

How they're used in practice

Experienced swing traders treat these oscillators as a secondary confirmation, never a primary signal. The workflow is: (1) trend filter decides direction (only take longs in uptrends), (2) price structure decides location (entry near support/resistance/MA), (3) the oscillator decides timing (enter when momentum is stretched and turning back your way). The single most common professional convention is to flip thresholds with the trend: in an uptrend, treat 40 (RSI) / 20 (Stoch) as the buy zone and largely ignore overbought; in a downtrend, do the reverse. Multi-timeframe alignment — using a higher timeframe for trend and a lower one for the oscillator trigger — is the other standard refinement.

Adoption, debate & evidence

Adoption is near-universal; RSI and Stochastics ship as defaults in essentially every charting platform. The evidence is genuinely split. Standalone overbought/oversold rules have weak-to-no edge — ChartSchool itself notes RSI's OB/OS readings "work best when prices move sideways within a range," that in strong trends RSI can stay overbought or oversold for extended periods, and that "divergences are misleading in a strong trend" (bearish divergences appear repeatedly in uptrends without producing reversals). The strongest documented results come from short-period RSI mean reversion with a trend filter: Connors/Alvarez and independent replications (e.g. QuantifiedStrategies RSI-2) report high win rates (commonly cited 70–85% on US index ETFs) — but these are short-hold, small-average-win systems vulnerable to occasional large losses, and high win rate is not the same as high expectancy. Treat any single backtest win-rate figure skeptically: they are sensitive to instrument (works far better on mean-reverting indices than on individual momentum stocks), period, and survivorship.

Honest separation of look-alikes: the academic momentum factor (Jegadeesh–Titman, 1993 — buying 3–12 month relative winners) is a robust, peer-reviewed cross-sectional anomaly. The RSI indicator is an unrelated short-term bounded oscillator with weak standalone evidence. They share the word "momentum" and nothing else; RSI does not inherit the factor's academic credibility.

Strengths & limitations

Strengths: excellent at flagging short-term exhaustion for entry timing inside an already-validated setup; intuitive, fast to read, and effective for the multi-day reversion swings these tools were built for. RSI(2)-style reversion has the most replicated edge.

Limitations: no directional information on their own; they lag at major turns and pin at extremes during strong trends — the #1 misuse is fading a trend purely on an overbought/oversold reading. They are also collinear (both momentum-derived), so stacking RSI + Stochastics is redundant confirmation, not independent confirmation.

Sources

Disputes flagged: (1) standalone OB/OS edge is weak/contested; (2) RSI-2 high win-rate claims are real but instrument- and regime-dependent and conflate win rate with expectancy; (3) "momentum" naming overlap with the academic factor is a common credibility-borrowing error.