What Is Williams %R?
Williams %R measures where the close sits inside the highest high and lowest low of the last 14 bars, on a scale from −100 to 0. It is arithmetically the same series as an unsmoothed stochastic %K with 100 subtracted from it.
Most pages here explain what an indicator measures. This one mostly explains that you may already have it.
How it works
Where did the close sit inside the last 14 bars?
%R = −100 × (highest high − close) / (highest high − lowest low)
Close at the top of that range and you get 0. Close at the bottom and you get −100. The negative scale is Larry Williams’s choice and carries no extra meaning.
It is the stochastic
This is the page. Compare the formula above with the stochastic page’s:
%K = 100 × (close − lowest low) / (highest high − lowest low)
The numerators are the same distance measured from opposite ends, so %R = %K − 100. Not
approximately — exactly.
Verified on this chart: across all 60 bars where both are defined, the largest difference between Williams %R and raw %K minus 100 is 0.000000000000.
So it is not a second tool, a confirmation, or a different perspective. It is the same number with 100 taken off it and the sign of the scale flipped.
The one real difference
Williams %R is conventionally plotted raw. The stochastic is conventionally smoothed.
TradingView’s stochastic defaults to 14, 3, 3 — the raw calculation averaged over 3 bars, then averaged again for the signal line. Williams %R has no smoothing parameter at all.
So the practical difference is responsiveness, not information. %R moves a bar or two sooner and is correspondingly noisier, and if you want that behaviour from the stochastic you set its smoothing to 1 and get this line back.
The zones
−20 and −80 are the conventional overbought and oversold lines, and they are the same 20% and 80% the stochastic uses, relabelled.
7 of those 10 bars were inside an advance. The same trap as the stochastic, arriving in the same place for the same reason: a market making higher highs closes near the top of its recent range by construction, so “overbought” is a description of an uptrend rather than a warning about one.
The settings
One number: the lookback, 14 by default.
Lengthening it widens the window the close is measured against, which makes new extremes rarer and the line slower. There is no length that fixes the trend problem above, because the problem is in what the calculation asks, not in how many bars it asks it over.
Why two identical indicators both exist
Worth a section, because the answer generalises well past this page.
They were invented separately. George Lane’s stochastic and Larry Williams’s %R come from the 1950s and 1973 respectively, developed independently on the same observation — that where a bar closes inside its recent range says something about who was winning.
Neither author knew the other’s work was arithmetically the same, and by the time anyone noticed, both had names, followings and books.
Then platforms shipped both, because a charting package competes on how many indicators it has, not on how many distinct measurements it offers.
The general lesson is the useful part. Your platform’s indicator list is a historical accident — an accumulation of everything anyone ever published, deduplicated by nobody. Two entries in that list are not two measurements, and the only way to know is to read the formulas.
That is not an argument against either tool. It is an argument against the feeling of confirmation you get when two of them agree.
A worked example
Establish that you are in a trading range first, from market structure. This tool is built for a market that returns to the middle.
Mark the range edges from price. Those are the levels; the panel is not.
Then a reading above −20 at the top of a real range is a second reason, not a first one.
And the invalidation is a close outside the range — a price, from the chart. Nothing in a −100 to 0 scale can give you one.
The original data
Across our study of 24,971 trading videos, 52 cover Williams %R. The median one gets 3,766 views, 88% never pass 50,000, and the median length is 6.8 minutes.
6.8 minutes is among the shortest medians measured here, against 9.6 for commodity channel index (CCI) and 9.2 for average directional index (ADX) — which is roughly what a tool with one formula and one setting deserves.
The corpus carries description text for 47 of those 52, and across those 47, one mentions invalidation, failure, or what a bad read looks like.
When it fails
It fires constantly in the condition it suits
Six zone entries in 42 sideways bars. Being unsmoothed is the cause: every push toward the range high takes it above −20, including the ones that mean nothing.
This is the cost of the responsiveness, and it is the reason the smoothed stochastic became the more popular of the two identical tools.
In a trend it is wrong for as long as the trend lasts
Covered above with numbers. Seven of ten overbought bars were inside a rise. No setting changes that.
You are running it alongside the stochastic
Then you have one indicator twice, and when they agree — which is always, because they are the same series — it feels like confirmation.
You read the zone after the turn
Pinned at the top before a reversal and pinned at the top in the middle of a strong trend are the same picture, and this page’s whole subject is a tool that cannot tell you which one you are in.
Related
Stochastic is the same calculation with smoothing and a 0–100 scale, and the page that covers what both are actually measuring.
Trading range is the condition either of them needs.
And relative strength index (RSI) is the oscillator that genuinely measures something different — the balance of up moves to down moves rather than position in a range.
I found out these were the same series by accident, by having both on a chart and noticing one was a mirror of the other. That was a useful few minutes because it made me go and check what else I was running twice. If you are going to use it, the honest reason is that you prefer the unsmoothed version, not that it is a second opinion - it cannot be a second opinion on a calculation it is identical to.
— Michael Whitman, from this video
This page is educational, not financial advice. Test every idea on your own charts before risking money.