IV Rank vs IV Percentile: Why They Disagree

On 24 August 2026, Johnson & Johnson's at-the-money implied volatility was 32.57%. Measured against 1,288 days of its own stored history, that reading was IV rank 30 and IV percentile 92.

Same ticker. Same day. Same history. One statistic says the volatility being priced sits in the bottom third of where it has been; the other says it sits above 92% of every session on record. Neither number is a mistake, and neither is rounded, misreported, or stale. They disagree by 62 points because they are not measuring the same thing — and on the IV Percentile Screener they sit in adjacent columns, which is exactly where that difference is easiest to miss.

This is what each one computes, why the gap between them is structural rather than occasional, and how large it actually is across the whole universe.

The two formulas, as they are implemented

Both are computed in the same function, from the same list of stored daily readings, at the same moment:

IV rank places the current reading inside the range of the history.

IV rank = (current − historic low) / (historic high − historic low) × 100

IV percentile counts how much of the history sits below the current reading.

IV percentile = (days with IV < current) / (total days) × 100

Read them side by side and the asymmetry is already visible. IV rank consults exactly two observations out of the entire history — the highest and the lowest — and ignores every other day. IV percentile consults all of them and ignores their magnitudes entirely; it only asks, for each day, whether it was below today.

Rank is a range statistic. Percentile is a distribution statistic. Everything that follows comes from that one distinction.

Why the gap is structural, not occasional

Implied volatility does not spread itself evenly across its range. It spends most of its life in a low band and occasionally spikes hard, which stretches the historic high far above the typical day and leaves the denominator of the rank formula describing a range the series almost never visits.

That is measurable directly. For each of the 2,778 tickers with at least 30 stored sessions, take the median day of its own history and ask where that median sits inside its own high–low range. If volatility were spread evenly, the answer would be about 50.

The answer is 15.7.

And it is not a handful of names pulling an average around: 2,765 of 2,778 tickers — 99.5% — have their median session sitting below the midpoint of their own range. The typical ticker's mean IV is 1.12 times its median, the signature of a right-skewed series.

So the median day of the median ticker scores about 15 on rank and, by construction, exactly 50 on percentile. The gap is not an artifact of a strange market week. It is what these two formulas do to a right-skewed series, every day, by design.

Johnson & Johnson is the clean illustration. Its stored history runs from a low of 8.30% to a high of 88.64% — a range of 80 points — but its median session is 22.63%, sitting just 17.8% of the way up that range. Most of its days are crushed into the bottom fifth. A reading of 32.57% clears the great majority of them (percentile 92) while still sitting low inside a range whose top end was set by a handful of spike days (rank 30).

How far apart they actually run

Measured across all 2,778 tickers with sufficient history, comparing each one's rank against its percentile on the same day:

Half the universe disagrees with itself by more than 20 points. Nearly a quarter disagrees by more than 40.

The gap also has a direction. IV percentile is the higher of the two on 2,468 of 2,778 tickers — 88.8%. Rank is higher on 241, and the two are exactly equal on 69, all of which turn out to be the same case: a ticker sitting precisely at its historic low, where rank is 0 and percentile is 0 too.

That directional bias is the part worth carrying away. If you are used to reading one of these numbers and you switch to a screen showing the other, you are not swapping in a noisier version of the same quantity — you are systematically shifting your sense of where volatility sits, in a predictable direction, by roughly 20 points.

What that does to a threshold

Thresholds are where an abstract 24-point gap turns into different screen output. Applying the same cutoffs to both statistics across the same 2,778 tickers on the same day:

A rule written as "60 or above" selects 33 names when applied to rank and 688 when applied to percentile — a factor of twenty-one, from the same underlying volatility, on the same day. In the other direction, a "20 or below" rule captures 76% of the universe on rank and 32% on percentile.

And 244 tickers manage both at once: rank at or below 20 while percentile is at or above 60. On those names the two columns place the same volatility reading in the bottom fifth and the top two-fifths simultaneously.

Neither cutoff is wrong. A threshold is only meaningful attached to the statistic it was calibrated on, and the two statistics have different distributions, so a number that means "unusually high" on one means something else entirely on the other.

Ten names, same day, both columns

Amazon is the instructive row at the bottom: a gap of 0.8 points, the two statistics in near-perfect agreement. Agreement happens — it is simply not the default, and there is no way to tell from a single row whether you are looking at an Amazon or a Johnson & Johnson without knowing what the underlying distribution looks like.

Tesla is worth a second look too. Its 67.00% reading is the highest raw implied volatility in the table by a wide margin, and it carries the lowest IV rank in the table at 10.7. High absolute volatility and a low rank are entirely compatible: rank says nothing about the level, only about the position inside that ticker's own history, and Tesla's history reaches 260%.

The floor, the clamp, and the tie

Three implementation details change what you see, and all three are deliberate.

A 30-observation floor. Below 30 stored sessions, both statistics return nothing rather than a number. This was raised from 5 after a review found that 382 of 1,252 tickers — 31% — were being classified into volatility regimes on histories short enough that a single new observation could swing the rank by 20 points or more. A range statistic computed from five samples is an estimate of two extremes from five draws, which is close to no information at all. Surfacing "limited IV history" beats surfacing a confident-looking number built on noise.

A clamp at 0 and 100. Current IV can exceed the historic high — that is what a new high is — which would otherwise produce a rank above 100. Eight tickers in the current universe sit at exactly 100 for this reason. The clamp keeps the metric inside the range its name implies, at the cost of flattening every new high to the same value.

Ties count as below. The percentile comparison is strict: a stored day exactly equal to today's reading is not counted as below it. On a ticker whose IV has not moved, percentile reads lower than intuition suggests. It affects a small minority of tickers, but it is the reason a percentile of exactly 0 does not always mean "the lowest reading ever seen".

There is also one asymmetry between the two. If a ticker's history is perfectly flat — high equal to low — the rank formula divides by zero and returns nothing, while the percentile formula returns 0 quite happily. A missing rank next to a present percentile is not a data gap; it is a history with no range in it.

Reading them together

The practical version is short.

IV rank answers: where does today sit between the calmest and the wildest this name has ever been? It is sensitive to two historic extremes and therefore to single events years ago that may never repeat.

IV percentile answers: how many of this name's sessions were quieter than today? It is insensitive to magnitude, so a reading far above anything previously seen and a reading barely above the old high both score near 100.

Because they answer different questions, the interesting cases are the ones where they disagree loudly. A low rank beside a high percentile — the 244-ticker case above — describes a name that is elevated relative to its ordinary sessions but still nowhere near its own crisis levels. That is a genuinely different situation from a name where both read high, and the only way to see the difference is to look at both columns rather than picking one.

Equity Rank's options engine uses both, weighted separately, precisely because they are not substitutes: IV rank carries a 0.22 weight and IV percentile 0.13 inside the composite volatility reading, alongside the variance risk premium at 0.30 and term structure, skew, earnings proximity and liquidity making up the rest. Two statistics that agreed with each other would not both earn a weight.

Frequently Asked Questions

What is the difference between IV rank and IV percentile?

IV rank measures where the current implied volatility sits between the lowest and highest readings in the stored history, so it depends on just two historic observations. IV percentile measures what share of all historic days had a lower reading than today, so it depends on every observation but ignores how far apart they are. One is a range statistic, the other a distribution statistic.

Why is IV rank usually lower than IV percentile?

Because implied volatility is right-skewed. Occasional spikes push the historic high far above the typical session, which stretches the range that IV rank divides by. Measured across 2,778 tickers, the median session sits only 15.7% of the way up its own high–low range, and 99.5% of tickers have their median below the midpoint. IV percentile came out higher than IV rank on 88.8% of them.

How much do IV rank and IV percentile actually differ?

Measured on the same day across 2,778 tickers with at least 30 sessions of history: a mean absolute gap of 23.92 points, a median of 20.25, and a 90th percentile of 52.66. Half the universe differed by more than 20 points and just under a quarter by more than 40. The largest single gap was 78.11 points.

Can a stock have high implied volatility and a low IV rank?

Yes, and it is common. IV rank describes position within one ticker's own history, not the level of volatility itself. In the sample above, Tesla carried the highest raw implied volatility at 67% and the lowest IV rank at 10.7, because its stored history reaches far higher still.

Why do some tickers show no IV rank at all?

Two cases. Below 30 stored sessions both statistics are withheld, because a range computed from a handful of samples moves violently on each new observation. Separately, if a history is perfectly flat, the rank formula has a zero range to divide by and returns nothing while the percentile still computes.

Which one should a screener threshold be set on?

Whichever one it was calibrated on — the two are not interchangeable at a given cutoff. Applying a "60 or above" rule to the same universe on the same day selected 33 tickers on IV rank and 688 on IV percentile. A threshold carried over from one statistic to the other silently changes what it selects.

Related reading: IV Rank Explained · Implied Volatility Explained · What an Earnings Implied Move Actually Measures · Put-Call Parity Explained · Options Greeks Explained

Figures measured on 5 September 2026 against the stored daily implied volatility history, covering 2,778 tickers with at least 30 sessions each. Equity Rank publishes research tools and market data. It is not a registered investment adviser, and nothing here is investment advice or a recommendation regarding any security.