Williams %R shows where the latest close sits within the range of a recent window. A close at the highest high gives 0; a close at the lowest low gives −100. It does not measure whether price is expensive or cheap, and it does not predict the next bar by itself.
In plain terms — Picture a lift with floor 0 at the top and floor −100 at the bottom. %R tells you where price stopped: near 0 means near the recent high; near −100 means near the recent low.
From price to number
For an n-period window, let Hn be the highest high, Ln the lowest low, and Ct the current close:
%R = −100 × (Hn − Ct) / (Hn − Ln)
Example: high 65, low 55, close 58. The distance from the high is 7 and the range is 10, so %R is −70. The minus sign sets the direction of the scale: as the close falls towards the low, the value approaches −100.
The historical excerpt reproduced by the author's site calculates the original measure over ten days and expresses the distance from the high on a positive 0-to-100 scale. TradingView plots the same distance with a negative sign and uses 14 periods by default; other platforms may adopt different conventions. Ten and fourteen are therefore not mathematical laws: the window belongs to the indicator's specification.
Reading it without reversing it
The 0/−20 and −80/−100 bands are widely used references for closes packed respectively near the upper and lower parts of the range. They do not mean “sell” and “buy.” In a strong trend, %R can remain extreme because closes keep gathering near the same edge.
The −50 midpoint is more literal: above −50, the close lies in the upper half of the window; below −50, it lies in the lower half. This too describes position, not future direction.
A testable procedure
- State the window and data. Record bar count, timeframe, market, and price source before viewing the result.
- Read price first. Separate ranges from trends; the same extreme may mean different things in those regimes.
- Define the event. “%R below −80” is a state; a rule must specify crossing, re-entry, confirmation, and invalidation.
- Test out of sample. A backtest must include costs, repeated signals, and periods when an extreme persists without a reversal.
Its relationship with the Stochastic
With the same window and no smoothing, Williams %R and the fast %K line of the Stochastic contain the same information on shifted scales:
Williams %R = fast %K − 100
A close at the high gives %K = 100 and %R = 0; a close at the low gives %K = 0 and %R = −100. Combining them does not create independent confirmation. Differences on a chart usually come from the lookback, smoothing, price source, or platform rules.
Limits and implementation cases
- If Hn = Ln, the denominator is zero and the value is mathematically undefined. A platform should disclose whether it returns missing data, carries the prior value, or applies another convention.
- The current bar's %R can change until the bar closes; comparing an intrabar reading with a close-based test changes the rule.
- Sessions, gaps, historical adjustments, and different feeds can change the window's high and low.
- Optimising period and thresholds on the same sample used to judge performance creates overfitting risk.
Provenance and attribution
Larry Williams dates the development of Percent R to 1966 and acknowledges influence from Allan C. Davis. The book excerpt, which Williams says he wrote in 1973, documents the ten-day calculation and stresses that the indicator should be read within a market regime. This is an authorial primary source: it documents what Williams published and claimed, not an independent validation of the examples' profitability.
Sources
- Larry Williams, Original Williams %R — stated excerpt from the book written in 1973, ten-day definition, worked example, and limits described by the author.
- Larry Williams, Trading Indicators — claimed 1966 chronology and Allan C. Davis influence.
- TradingView, Williams %R — negative-scale formula, configurable levels, and that platform's default.
- Fidelity, Williams %R — mathematical relation to the fast Stochastic and persistence of extremes.
Related entries
Larry Williams · Stochastic · Price range · Trend · Backtest · Indicators