RSI vs Williams %R
The relative strength index compares recent gains with recent losses over a chosen length. Williams percent R measures how far the latest close sits below the recent high, which is the stochastic calculation on an inverted scale rather than a genuinely different idea.
Two bounded oscillators read for the same things. One of them is far better known and far better supported, and the other is a rearrangement of a third tool most people already have on the chart.
What each one is
The relative strength index compares recent gains with recent losses over a chosen length, on a scale from nought to one hundred. The relative strength index covers it.
Williams percent R measures how far the close sits below the recent high, expressed as a negative number from nought at the top to minus one hundred at the bottom. Williams percent R covers it.
That second calculation is the stochastic inverted. Both ask where the close sits within the recent range; one counts up from the low and the other counts down from the high.
Where they differ
What is being compared. Gains against losses in one; the close against the recent high in the other. Those produce similar-looking readings for different reasons.
Which way the scale runs. One reads high when the market has been strong; the other reads near zero. Getting that backwards is the single most common error with the second tool.
How much smoothing there is. The strength index averages gains and losses. Williams percent R applies none at all, which makes it the twitchiest reading of the family.
How well supported each is. One is everywhere, in every description of every method. The other is available and rarely discussed, so you will be working it out alone.
Where they agree
Both are bounded. A reading means the same shape of thing on any instrument, which is the shared advantage over anything with an open scale.
Both pin in a trend. A sustained move holds either at an extreme for many bars, and reading that as exhaustion is the standard error with both.
Both fail in a range. On this site’s shared series direction runs average 2.01 bars with a longest of 11, and short runs cross any threshold repeatedly on either.
And both cost a round trip per signal acted on — about 2% of the median bar range of 0.493 here — which the unsmoothed one pays more often because it produces more signals.
Which one to use
Run the relative strength index. It is smoothed, it is supported everywhere, and every strategy description you will read assumes it — which matters more than any property of the formulas.
Run Williams percent R when you want the rawest reading available. No smoothing means it turns first, and it turns first on everything.
Run it if you already read the stochastic and prefer the inverted display. That is a legitimate preference and it is honestly what the choice amounts to.
And never run it alongside the stochastic. Those two are the same measurement with the scale reversed, so having both is one number displayed twice, upside down.
Why the inversion causes trouble
Because every other oscillator counts the other way. A reading near zero means strength here and weakness almost everywhere else, and that reversal is easy to forget under pressure.
And because the thresholds sound the same. The familiar numbers appear in both tools with opposite meanings, so a rule copied from one to the other is exactly backwards.
What the lack of smoothing costs
More signals, all of them earlier. Every one arrives before the smoothed version’s, including the ones that lead nowhere, which is the same trade any faster tool makes.
Less tolerance for a single odd bar. One wide-range bar moves an unsmoothed reading to a limit, and on this site’s shared series the largest single bar range was 2.338 against a median of 0.493.
A higher trade count for the same idea. If your rule is a threshold cross, the unsmoothed version will cross it substantially more often, and each crossing costs the same round trip.
And no way to tune it down. The smoothed alternative has a length you can lengthen; the raw one does not, so if it is too noisy for you there is nothing to adjust.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, no title compares these two
directly — this pair is constructed from two subjects the corpus covers separately. Separately,
Williams percent R appears in 23 titles at a median of 6,222 across 23 channels, and the relative
strength index in 820 at a median of 5,021 across 529. The counts come from site/corpus_count.py.
23 videos across 23 channels — one each, and the thinnest coverage of any oscillator measured here. Nobody makes a second video about it, which is itself the practical argument: a tool with no community around it is one you will be debugging by yourself.
The answer to the question on that chart is strong, and that is the trap. Near zero means the close is at the top of its recent range — the opposite of what the same number means on every other oscillator you have used.
When it fails
The failure is copying a threshold rule across from another oscillator, and it inverts your entire method. A rule that says fade readings above a level is written for a scale that counts up from weakness. Applied to a scale that counts down from strength, the same rule fades exactly the wrong extreme. Every trade is taken at the opposite end of the move from where it was intended, and the chart looks normal throughout because both tools produce a familiar-looking line.
The second failure is running it beside the stochastic. They are one measurement.
A third is treating a limit as exhaustion. It reports one-sided bars.
A fourth is trading every cross on an unsmoothed reading. There are many.
A fifth is expecting a length to smooth it. There is nothing to lengthen.
And a sixth is adopting a tool with no community. You will be alone with it.
Related
The relative strength index covers gains against losses. Williams percent R covers the inverted range reading. And stochastic covers the calculation the second one is a presentation of.
The thing nobody says about Williams percent R is that it is the stochastic with the scale flipped. Once you know that, the choice is not between two ideas — it is between a widely supported tool and a rarely discussed presentation of one you already have.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.