Momentum oscillators measure the speed and extent of price change rather than direction. They are bounded (RSI and Stochastics 0–100; CCI is unbounded but calibrated around ±100) and answer three questions: is a move overextended, is momentum fading ahead of price, and where within a larger trend is a low-risk entry. Their defining weakness is the mirror of their purpose — because they measure how stretched a move is, they fire premature reversal signals in strong trends, where an oscillator can sit "overbought" for months while a leader keeps climbing.

How they're calculated

RSI (Wilder, 1978) — RSI = 100 − (100 / (1 + RS)), where RS = average gain / average loss. The averaging is specific: the first average gain/loss is a simple mean over the first 14 periods (the default); each subsequent value is smoothed as [(prior avg × 13) + current] / 14. This is an exponential average with smoothing factor α = 1/n — slower than a standard EMA's α = 2/(n+1), so RSI is steadier than an EMA of the same period.

Stochastics (Lane, ~1957) — measures where the close sits within the recent range: %K = (close − lowest low) / (highest high − lowest low) × 100, with %D = 3-period SMA of %K (the signal line). Default lookback 14. Fast stochastics use raw %K (noisy); Slow stochastics (14, 3, 3) smooth %K once more and are the practical default.

CCI (Lambert, 1980) — CCI = (TP − SMA(TP)) / (0.015 × mean absolute deviation), where typical price TP = (High + Low + Close) / 3, default period 20. The 0.015 constant is calibrated so that roughly 70–80% of readings fall within ±100 (making CCI comparable across instruments). Normalizing by mean absolute deviation (not standard deviation) makes it more robust to outliers.

How to read them

Crucially, an extreme reading is not a signal by itself. Wilder's own framing was that RSI > 70 means a reaction is likely, not imminent — a distinction modern usage often loses. The more durable use of RSI is as a trend filter: in a healthy uptrend RSI holds above ~40 and tags 70+; a trend's character has changed when it can no longer reach its usual extreme.

How it's used in practice

These oscillators get deployed a handful of recognized ways — none of them "buy the extreme blindly":

• Mean-reversion in ranges: fade overbought/oversold back toward the middle. Effective only when price is range-bound; this is the use that gets traders killed in a trend.
• Trend confirmation / filter: require the oscillator to agree with the trend rather than fade it — RSI holding above ~40–50 in an uptrend, or a centerline (50) cross for directional bias.
• Pullback timing: inside an established uptrend, a mild oscillator dip (e.g. RSI easing to ~40–50) marks a lower-risk entry than chasing extended strength.
• Short-term mean reversion — RSI(2): Larry Connors' method — buy a washout (RSI(2) < 5) within an uptrend and exit into strength. A fast, distinct use from the 14-period read.
• Divergence (next section) — to anticipate reversals or confirm continuation.
• Confluence: as a secondary confirmation alongside price structure, support/resistance, and volume — never the sole trigger.

These are the general approaches. The swing-specific operationalization — the exact entry, stop, target, and hold period for, say, an RSI(2) bounce — lives in Swing Trading → Indicators in a Swing Context.

Divergence — the highest-value read

• Regular divergence (reversal): price makes a higher high but the oscillator makes a lower high (bearish, in an uptrend); or price a lower low while the oscillator makes a higher low (bullish, in a downtrend). Momentum is fading ahead of price.
• Hidden divergence (continuation): price a higher low but oscillator a lower low (bullish, uptrend); or price a lower high but oscillator a higher high (bearish, downtrend). Signals the trend resumes.

Divergence is suggestive, not deterministic — it can persist across several swings before resolving, so pair it with a price trigger.

Worked example — RSI(14)

Over 14 periods, suppose total gains sum to 14 and total losses to 4. First average gain = 14/14 = 1.0; first average loss = 4/14 ≈ 0.286. Then RS = 1.0 / 0.286 ≈ 3.50, and RSI = 100 − (100 / (1 + 3.50)) = 100 − 22.2 = 77.8 → overbought. (In a strong uptrend that 77.8 is confirmation of strength, not a reason to sell — see below.)

Adoption, debate & evidence

Adoption. RSI, Stochastics, and CCI are standard built-in defaults on essentially every charting platform (TradingView, ThinkorSwim) and are among the most widely used tools in retail trading. Professionals use them too — but as context and confirmation, never standalone; the discipline is to layer them with trend and volume. Systematic/quant funds lean on the momentum factor (below), but treat oscillator indicators as, at most, minor features inside multi-factor models — not primary signals.

Contestation. Whether these indicators actually predict is genuinely disputed. The weak-form Efficient Market Hypothesis holds past prices are already discounted, making indicator-based prediction futile; decades of academic work are mixed and plagued by data-snooping — one review found 56 of 95 technical-analysis studies positive, but many contaminated by overfitting. The overbought/oversold-reversion premise is the most contested piece, since markets routinely stay overbought or oversold for weeks in a strong trend.

Evidence — the distinction that matters most. Do not conflate two different things that share the word "momentum":

• The momentum factor (Jegadeesh & Titman, 1993 — buy 3–12-month winners, sell losers) is one of the most robust documented anomalies in finance (~1.3%/month on the strongest specification, replicated for 30+ years, used by quants). Real.
• The oscillator indicators (RSI/Stochastics/CCI) have no robustly documented standalone edge. Backtests range from outright losses to modest wins (≈58–62% hit rates with small average winners), improving only once filters or a second signal are added.

The factor being real does not mean an RSI overbought reading predicts a reversal — different phenomena. Honest bottom line: treat these as diagnostic/contextual tools (they accurately describe momentum and exhaustion, best in ranges), not predictive signals.

Strengths & limitations

• Strength: an early, objective read on fading momentum; computable on any series; divergence often precedes price turns.
• The #1 misuse: shorting "overbought" in an uptrend. RSI stayed overbought for months while NVIDIA rose ~550% across 2023–2024. In strong trends, widen thresholds (e.g. 80/20, or 80/40 up- / 60/20 down-trend) or lean on the trend-filter read instead.
• Per-oscillator: Stochastics whipsaws badly in trends and chop — prefer Slow (14,3,3) and trade with the trend. CCI is built for cyclical/trending moves and whipsaws near the zero line in ranges; CCI(30) reduces false signals at the cost of lag.
• Universal: no oscillator is reliable in isolation or against the prevailing regime; use them as confirmation, gated by price structure and trend.

Sources

• StockCharts ChartSchool — RSI, Stochastic Oscillator, CCI (formulas, defaults, thresholds)
• J. Welles Wilder, New Concepts in Technical Trading Systems (1978) — RSI / Wilder smoothing
• CMT Association — origins of the Stochastic Oscillator (George Lane)
• Donald Lambert, Commodities magazine (Oct 1980) — CCI and the 0.015 constant
• Investopedia / Wikipedia — cross-checks on originators, periods, ±100 calibration
• Disputes flagged: CCI "% within ±100" cited as 70–80% (most sources) vs ~85% (some); RSI 70/30 as "reaction likely" (Wilder) vs "reversal imminent" (modern usage)
• Jegadeesh & Titman (1993), Returns to Buying Winners and Selling Losers — the momentum factor (distinct from the oscillators)
• CFA Institute — the Efficient Market Hypothesis and its critics; Han/Liu/Zhou/Zhu — technical-analysis review (data-snooping; 56/95 studies positive)
• Quantpedia — momentum factor effect; QuantifiedStrategies / StockBros Research — RSI strategy backtests (mixed/weak standalone results)
• Adoption: TradingView & ThinkorSwim default indicator sets (cross-checked platform comparisons)