Williams %R indicator: settings and signals
Williams %R shows how far the closing price is from the high of the recent range, on an inverted scale from −100 to 0.
What the indicator looks like on a real chart
Data: EUR/USD, daily bars, 2025-06-16 — 2026-07-20. The indicator is computed over 400 bars; the chart shows the last 180. Quote source: Binance Spot REST API (api.binance.com/api/v3/klines).
| Regime | Score |
|---|---|
| Trending market | 40 |
| Range | 62 |
| High volatility | 38 |
| Low volatility | 53 |
What the indicator actually measures
The calculation almost coincides with the Stochastic, but the reference point differs: it measures the distance not from the low upward but from the high downward. Hence the inverted scale, where zero at the top means a close at the period high, and −100 at the bottom a close at the low.
Mathematically %R and the Stochastic %K are linked by a simple relation: the absolute values of the two sum to 100. In practice this means the charts of the two indicators are mirror images, and using both at once as independent confirmations is meaningless.
Larry Williams proposed a reading different from the customary one: he regarded reaching extreme values in the direction of the trend not as a reversal signal but as a sign of strength. This is a rare case where the author's reading of an indicator diverges from the one that took hold in mass usage.
Formula
The numerator is the distance from the period high to the close, the denominator the entire period range, and the result is taken with a minus sign. Because of the negative scale, "overbought" is at the top of the chart rather than at the bottom, as with most oscillators.
Standard settings
| Parameter | Value |
|---|---|
| Standard period | 14 |
| Overbought / oversold | −20 / −80 |
| Period for scalping | 9 |
Period 14 coincides with the standard of the Stochastic and RSI. The levels −20 and −80 correspond to 80 and 20 on the Stochastic. There is no smoothing at all in the original formula, so the indicator is even sharper than the Stochastic, and in practice it is often additionally averaged.
Implementation code (Python)
def williams_r(high, low, close, period=14):highest = high.rolling(period).max()lowest = low.rolling(period).min()return -100 * (highest - close) / (highest - lowest)
When the indicator stops working
- It fully duplicates the information of the Stochastic — using them together adds no independent confirmation.
- The absence of smoothing makes it the noisiest indicator in the group of range oscillators.
- In a trend it pins to the edge of the scale for a long time, like all indicators based on position within a range.
Common mistakes in use
- Using %R together with the Stochastic while treating them as different indicators. It is the same quantity in different coordinates.
- Getting confused with the signs and reading −20 as oversold. Because of the inverted scale, overbought is at the top, near zero.
- Trading every touch of a bound without smoothing — raw %R touches the extremes very often.
- Ignoring the author's reading: Williams regarded reaching extreme values in the direction of the trend as confirmation of strength, not a reason for a reversal trade.
Who needs this and why
Makes sense for those more accustomed to an inverted scale and for short-term range trading. There is no point in adding it to a system that already has a Stochastic.
Frequently asked questions
How does Williams %R differ from the Stochastic?
Practically only in the coordinate system. %R measures the distance from the high downward and expresses it in negative numbers; the Stochastic measures from the low upward and in positive numbers. The sum of their absolute values is always 100, so in substance it is one and the same calculation.
Why is the scale negative?
That is how the author designed it: the period high is taken as the zero point, and a deviation downward from it is naturally expressed as a negative number. Because of this the overbought zone ends up at the top of the chart, which regularly confuses those used to ordinary oscillators.
Do I need %R if I already use the Stochastic?
No. They carry the same information, and having both in a system creates a false sense of a signal confirmed by two independent sources.