How to Use Williams %R
To use Williams percent R, read it as the position of the close within the recent high-low range. It runs from zero at the top of the range to negative one hundred at the bottom, and it is arithmetically the same measurement as the stochastic with the scale reversed.
Williams percent R answers one question: where in the recent high-low range did price close. It runs from zero at the top to negative one hundred at the bottom, and it is the stochastic’s raw calculation with the scale reversed.
Before you start
A decision about whether the market is ranging or trending, made before any reading. The same value means exhaustion in one and strength in the other, so this cannot be settled afterwards.
The lookback set from your holding period rather than left at a default. It defines what “recent” means, and nothing else in the calculation is adjustable.
An understanding that it is the stochastic inverted, so the two are one indicator. If both are on your chart, you have one measurement drawn twice.
The steps
1. Read it as position in the range
A reading near zero means the close was near the highest high of the lookback. It contains nothing about value, fairness or direction.
2. Classify the market first
In a trend, closes sit consistently near one end, so the reading stays extreme for many bars. Reading that as exhaustion is the error this indicator produces most.
3. Set the lookback from your horizon
Fourteen bars describes a fortnight on a daily chart. If you hold for a month, a fourteen-bar window is answering a question about a period shorter than your trade.
4. Require a structural reason alongside the reading
An extreme reading at a level price has already respected is a different event from one in open space, and the indicator cannot distinguish them.
5. Wait for the reading to leave the extreme
Being extreme is a condition that can last twenty bars. Crossing back out of it is a discrete event with a timestamp, which is the version you can act on.
6. Take the stop from the chart
Where the idea is wrong is structural. On this site’s shared series the ninetieth percentile bar range is 1.101, so a stop inside that band is not a level.
7. Do not run it alongside the stochastic
They are the same calculation. Two lines agreeing is persuasive and here it is arithmetic, which is the least informative kind of agreement available.
How to tell it worked
The market type was decided before any reading was interpreted.
0 trades came from an extreme reading alone, without structure.
Entries were taken on the reading leaving the extreme, not on it being there.
And the chart carries 1 range-position oscillator, not two versions of the same one.
What it cannot see
Anything outside its lookback window. The range it measures against is the last N bars and nothing else, so a level from three months ago is invisible to it.
Whether the range is meaningful. In a very tight market the denominator is tiny, so trivial movement produces readings at both ends within a few bars.
How it differs from the stochastic in practice
The raw calculation is identical, inverted. Zero here corresponds to one hundred there, and negative one hundred corresponds to zero.
The difference is smoothing. The stochastic as commonly displayed applies smoothing to produce two lines and a crossover; this is usually shown raw, which makes it faster and noisier.
Which means the honest choice is about smoothing, not about which indicator is better. If you want a crossover, use the smoothed version. If you want the raw position with no lag added, use this one.
The negative scale, and why it confuses people
The output runs from zero down to negative one hundred, which is the reverse of nearly every other oscillator in common use.
Zero is the strong end. A reading of negative five means the close was almost at the highest high of the window, which most people’s instincts read as weakness because the number is negative and small.
Negative ninety-five is the weak end, and the same instinct reads it as safer because the magnitude is larger.
Flipping the chart vertically removes the problem entirely. Most platforms will do it, and once the scale runs the familiar way the readings mean what they look like — which matters, because an indicator you have to translate before acting on is one you will translate wrongly under pressure.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 23 mention this indicator in the
title, at a median of 6,222 views across 23 channels, and 61% of those titles are instruction-shaped.
The stochastic appears in 184 at 11,915 and the relative strength index in 154 at 4,398. The counts come
from site/corpus_count.py and site/rank_howto.py.
23 videos against 184 for the same measurement on a different scale. An eighth of the coverage for arithmetically identical information, which is a reasonable illustration of how much of an indicator’s popularity is presentation.
The answer to the question on that chart is that nine bars near zero describes a trend. Closes have been at the top of the range every day, which is what an advancing market does — and on this site’s series direction runs average 2.01 bars with a longest of 11, so persistence is ordinary.
When it fails
The failure is a trend read with range assumptions, and it repeats until the account notices. The reading pins near zero, every bar looks like a better short, and price keeps advancing. Each reading is accurate about what it measured — closes near the top of the recent range — and the question being asked of it, whether the move is finished, is one a position measurement has never been able to answer.
The second failure is running it with the stochastic. That is one indicator twice.
A third is acting on the condition rather than the crossing out of it. Extremes persist.
A fourth is a default lookback. It sets what the reading is about.
A fifth is comparing readings across instruments. It is scale-free by construction.
And a sixth is stopping from the oscillator. It contains no prices.
Related
Williams percent R covers the calculation. Stochastic is the same measurement smoothed and inverted. And trading range is the condition where the standard reading applies.
I stopped running this alongside the stochastic once I worked out they are the same calculation with the sign reversed. Having both on a chart looked like confirmation and was arithmetic repeating itself, which is the most convincing kind of false agreement there is.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.