Williams %R
An inverse stochastic oscillator ranging from -100 to 0, measuring where price closed relative to the period's high-low range.
Overview
Williams %R (Percent R) was developed by Larry Williams and published in 1973. It is essentially an inverted Stochastic Oscillator — instead of measuring where price closed relative to the low of the range, it measures how far below the period's high the close settled. The scale runs from -100 (at the very bottom of the range) to 0 (at the very top). Overbought is -20 and above; oversold is -80 and below.
How it looks on a chart
Illustration only — synthetic data generated for visual reference.
Williams %R works much like the Stochastic Oscillator, but the scale is flipped. A reading near 0 means price closed very close to the highest point of the period — strong recent buying. A reading near -100 means price closed near the lowest point — strong recent selling. Overall, readings above -20 are overbought and suggest the asset may be due for a pullback. Readings below -80 are oversold and suggest a possible bounce. These extremes are more significant when they occur alongside other confirming signals. One feature that distinguishes Williams %R from the Stochastic is that it is not smoothed by default. This makes it more sensitive and faster — it can identify reversals slightly earlier, but also generates more noise. Many traders smooth it with a 3-period moving average to reduce whipsaw.
The formula: %R = -100 × (Highest High₁₄ − Close) / (Highest High₁₄ − Lowest Low₁₄). The result is always negative (or zero). The key difference from %K in the Stochastic is the reference point: Stochastic measures distance from the low of the range, Williams %R measures distance from the high. Mathematically, %R = %K − 100. Larry Williams' original trading method used %R on weekly bars for stock selection: buy when %R drops below -90, wait for it to rise above -80 (confirmation), then enter long. This two-step method reduces false entries by requiring the oversold condition to be confirmed by a recovery before committing. For shorter-term trading, %R(5) or %R(7) on daily charts provides very sensitive signals. Combining Williams %R with a 200-period moving average as a trend filter — only taking oversold %R signals when above the MA — is a well-documented mean-reversion/trend-continuation hybrid approach.
Williams %R and the Stochastic Oscillator are mathematically equivalent (one is a linear transformation of the other), so their predictive information content is identical when both use the same lookback period and smoothing. The practical difference is that %R is displayed uninverted and without smoothing, making it appear more sensitive. In systematic factor research, rolling Williams %R over multiple periods (e.g., averaging %R over 5, 10, and 20 period lookbacks) creates a multi-scale momentum signal that is more robust than any single lookback period. This composite approach reduces parameter sensitivity — one of the main criticisms of momentum oscillators. Larry Williams himself used %R as part of a timing system for futures markets with significant published results. However, subsequent independent replications with out-of-sample data show more modest risk-adjusted returns, highlighting the general overfitting concern in technical indicator research. Gilito addresses this through strict walk-forward optimization on Williams %R parameter sweeps.
Formula
%R = -100 × (Highest High₁₄ − Close) / (Highest High₁₄ − Lowest Low₁₄)
- 1.Determine the lookback period n (default 14 periods).
- 2.Find the Highest High and Lowest Low over the last n periods.
- 3.Compute %R = -100 × (Highest High − Close) / (Highest High − Lowest Low).
- 4.Values range from -100 (close at period low) to 0 (close at period high).
- 5.Optionally smooth with a 3-period EMA to reduce noise.
Parameters
| Parameter | Default | Range | Description |
|---|---|---|---|
| Period | 14 | 3–50 | Lookback period for the high-low range calculation. |
| Overbought Level | -20 | -30–-5 | Level above which the asset is considered overbought (less negative = more extreme). |
| Oversold Level | -80 | -95–-70 | Level below which the asset is considered oversold. |
Trading signals
bullish: %R rises above -80 after being below -80
Oversold recovery confirmation — potential reversal long entry.
bearish: %R drops below -20 after being above -20
Overbought breakdown confirmation — potential reversal short entry.
bullish: %R persistently near 0 in an uptrend
Price consistently closing at highs — strong bullish trend continuation.
bearish: %R persistently near -100 in a downtrend
Price consistently closing at lows — strong bearish trend continuation.
Limitations
- •Mathematically identical to the Stochastic %K; no unique information versus Stochastic.
- •Uninverted scale (-100 to 0) is counterintuitive for many traders.
- •Like all overbought/oversold indicators, fails in strongly trending markets.
- •No smoothing by default makes raw %R noisy and prone to whipsaw signals.
Gilito tests Williams %R across all standard lookback periods (5, 7, 10, 14, 20) and compares its signal quality to RSI and Stochastic on the same asset/timeframe combinations to identify the most effective short-term momentum oscillator for each instrument class. Multi-period composite %R signals are also backtested as a more robust alternative.