Direct answer
Williams %R behaves differently across market conditions mainly because it normalizes the current close within the recent high–low range. When that recent range is unusually narrow or wide, or when price is trending versus oscillating, the normalized value can become more extreme or more unstable—even if the raw price movement looks similar.
It is important to separate (1) the indicator’s stable computation rule from (2) variable market features such as volatility, trend strength, and how often price reverses inside the lookback window.
Mechanism or definition
Williams %R (often written %R) is a momentum oscillator computed from a lookback period (commonly N). Conceptually, it answers: “Where is today’s closing price relative to the highest high and lowest low over the last N periods?”
A typical formulation uses the lowest low and highest high over the lookback window and compares the current close to that range. The result is a bounded oscillator (commonly expressed from about −100 to 0). The key mechanics are stable: the indicator rescales price into a percentage-of-range measure.
Because the numerator is tied to the close’s position, and the denominator is the recent range (highest high minus lowest low), behavior changes when the market’s recent range changes. For example:
- In a compressed range, small changes in close can move %R substantially.
- In an expanded range, the same absolute move in close can translate to a smaller change in %R.
Evidence or example
Below are condition-based scenarios that change Williams %R readings, expressed with explicit assumptions.
Scenario A: Narrow recent range (low realized volatility) Assume N is fixed. Suppose over the last N periods, the highest high is very close to the lowest low. If today’s close shifts slightly toward the top of that narrow band, the close’s “distance from the low” relative to the small range becomes large. Williams %R therefore moves toward its upper extreme more quickly.
Scenario B: Wide recent range (high realized volatility) Assume the same N. If the last N periods include a large high–low spread, the denominator is larger. Even if price moves sharply in absolute terms, the normalized position of the close may change less, so %R can appear to move more slowly or remain less “extreme” than a reader expects from the raw chart.
Scenario C: Trending market versus choppy market Assume the price movement stays near one side of the lookback range. In a sustained upward trend, the highest high within the lookback window tends to keep getting updated, and closes often remain closer to the top of the range; %R can therefore spend more time near an extreme. In contrast, in a sideways or mean-reverting environment, repeated reversals keep refreshing both ends of the lookback window; %R can swing back and forth more frequently or appear to “stall” around middle values.
Scenario D: Lookback window edges Assume N is fixed but the volatility regime changes right as the window boundary moves. When the oldest prices drop out of the lookback and a different set of highs/lows enters, the denominator can jump. That can cause a visible “regime shift” in the oscillator even without a major change in today’s raw direction.
Limitations and risks
-
Indicator does not predict future direction. Williams %R measures relative position inside a historical window. Past normalization and historical extremes do not guarantee future follow-through.
-
Edge cases can create misleading stability or volatility. If the recent high–low range is extremely small, %R can become highly sensitive to tiny price changes. Conversely, if the market’s range structure remains similar, %R may look stable while conditions beneath the surface evolve.
-
Parameter and data assumptions matter. The behavior depends on the lookback period N and on which prices are used (high, low, and the chosen “close” concept). Different data feeds, different candle construction, or different N can change the oscillator’s appearance.
-
Execution and costs can dominate outcomes. Even though this article is informational, readers should note that market microstructure, bid–ask spreads, slippage, and jurisdiction-specific rules can affect how any indicator-based analysis translates to real results. Those factors are external to the mathematics of Williams %R.