Stochastic oscillator

Definition

The stochastic oscillator (George Lane’s %K/%D construction) locates the current close within the high-low range of the last N bars, scaled 0-100: near 100 means closing at the top of the recent range, near 0 the bottom. It is a range-position gauge — closely related to RSI in use, but measuring position-in-range rather than gain/loss balance.

How it works / structure

• Formula: %K = 100 × (C − L_N) / (H_N − L_N), where L_N/H_N are the N-bar low/high; %D = SMA(3) of %K; the “slow” convention smooths %K itself first.
• Parameters (engine-executable): range window N (14 convention), smoothing windows, fast/slow variant, and the signal definition — threshold cross (20/80 convention), %K/%D crossover, or divergence (mechanically pinned).
• Relation to breakouts: %K = 100 means closing AT the N-bar high — the same event Donchian systems buy (indicator-donchian-channels); reversion readings of high stochastics and breakout entries are opposite interpretations of one event, resolved only by regime.

When it applies

Range-position triggers in range regimes (fade the edges); pullback timing within trends (slow stochastic re-crossing up from below 20 in an uptrend — a common swing convention, strategy-swing-trading); anywhere RSI-style logic is wanted with explicit range anchoring.

Risk profile & failure modes

• Pinning in trends: like RSI but more so — persistent trends close near range extremes bar after bar; edge-fading in trends is systematic loss.
• Degenerate ranges: flat N-bar ranges (H_N ≈ L_N) make the denominator vanish; quiet instruments produce noise spikes — the engine guards the division.
• Convention sprawl: fast/slow/full variants with three windows each — the same multiple-testing hazard as MACD, compounded by 20/80 threshold folklore.

Evidence & limits

Oscillator families show mixed, cost-fragile results in the surveyed literature (Park-Irwin 2007); no published study establishes standalone stochastic profitability after costs. The range-position quantity itself is descriptive fact; every signal convention layered on it is a hypothesis for replay.

Falsifiable-thesis examples

Illustrations only, not signals:

• “X, with slow stochastic re-crossing above 20 while its 200-day SMA rises, will close higher within 10 sessions” — falsified at the mark.
• “Fading %K > 95 on range-regime ETFs (exit at %K 50) is net profitable after friction this year in replay” — falsified by replay P&L.

Cross-references

• Closest sibling: indicator-rsi (gain/loss balance vs range position)
• The opposite reading of its extreme: indicator-donchian-channels, strategy-breakout
• Habitat: strategy-mean-reversion, strategy-swing-trading
• Method caveats: lens-technical