Skip to main content

IV Rank & IV Percentile

Updated Jun 24, 2026 at 2:35pm

Research Draft Medium 1,230 words

IV Rank and IV Percentile are two normalization tools that answer the same practical question — is the implied volatility on this underlying high or low right now, relative to its own recent history? — but they answer it differently. Raw implied volatility (e.g. "IV is 28%") is nearly useless on its own, because 28% is cheap for a biotech and expensive for a utility. Both metrics compress a year of IV readings onto a 0–100 scale so that one number is comparable across symbols and across time. The core tension between them is outlier sensitivity: IV Rank measures position within the year's range and is dominated by a single high print, while IV Percentile measures frequency below the current level and is robust to spikes. They frequently disagree, and the disagreement itself carries information.

How they're calculated

Both use a trailing one-year (≈252 trading day) window of the underlying's daily implied volatility, conventionally the at-the-money or a blended/30-day constant-maturity IV.

IV Rank locates today's IV between the year's extremes:

IV Rank = (Current IV − 52wk Low IV) / (52wk High IV − 52wk Low IV) × 100

Only two data points matter — the highest and lowest IV of the year. An IV Rank of 50 means current IV sits halfway between the floor and ceiling (StockCharts/Barchart conventions; Barchart and MenthorQ both publish this formula).

IV Percentile (also "IV Percentile of Days") counts how often IV was lower:

IV Percentile = (Number of days IV was below current IV) / (Total trading days) × 100

An IV Percentile of 80 means current IV is higher than it was on 80% of the past year's days. This uses all 252 observations, not just the extremes (Barchart, moomoo).

A worked example shows why they diverge: if IV spiked once to 90% during an earnings or crash event but has otherwise oscillated between 20% and 35%, and today's IV is 35%, IV Rank reads low (35 is near the bottom of the 20–90 range) while IV Percentile reads high (35 is above almost every other day). Because IV distributions are right-skewed — long stretches of low IV punctuated by sharp spikes — IV Percentile typically reads higher than IV Rank for the same underlying (Barchart).

How they're used in practice

The dominant use case is volatility-selling. Strategies that are net-short options (short strangles, iron condors, credit spreads, covered calls, cash-secured puts) profit when IV is rich and subsequently contracts. Practitioners use a high reading as a green light to sell premium and a low reading as a cue to buy options (long calls/puts, debit spreads, calendars) or to stand aside.

tastytrade popularized IV Rank as the default decision metric, with the commonly cited heuristic of deploying premium-selling strategies when IV Rank is above ~50, and a frequently quoted "high" threshold of 70+ (tastytrade is the original promoter of this framework; the specific cutoffs are house rules of thumb, not validated optima). Barchart cites a similar convention: both metrics above 70 = high vol, below 30 = low vol, 30–70 neutral.

The sophisticated move is to read the two together. When IV Rank and IV Percentile are both elevated, the high reading is "broad" — IV is genuinely sustained near its highs. When IV Rank is high but IV Percentile is low, the elevation is a transient spike (one event pulled the ceiling up), and several commentators argue IV Percentile is then the more trustworthy signal. A widely repeated rule is that a divergence of more than ~30 points almost always signals a single distorting spike (MenthorQ/FlashAlpha; this is an analysts' heuristic, not a measured statistic).

Adoption, debate & evidence

These metrics are near-universal in the retail and prop options world — tastytrade, Schwab/thinkorswim, Barchart, and most options platforms surface one or both. The deeper question is whether the underlying premise holds: does selling when "IV is high" actually pay?

The supporting evidence is the variance risk premium (VRP) — the robust empirical finding that option-implied volatility on average exceeds subsequently realized volatility. Carr & Wu (2009, Review of Financial Studies) documented a persistent, structurally negative variance risk premium across major equity indices over decades; it is widely characterized as positive (IV > RV) roughly 80–90% of the time on indices (commonly cited figures; the exact share depends on window and underlying). This is what makes systematic premium-selling profitable on average, and it is the genuine economic engine behind the IV-rank framework.

But two honest caveats separate the strong claim from the weak one. First, the VRP is an index-level, average phenomenon driven by hedging demand and skew/jump risk; it is far weaker and noisier on individual single names, where IV Rank is most commonly applied. The credibility of the academic VRP does not transfer automatically to "sell the high-IV-rank stock." Second, high IV rank is high for a reason — it usually reflects real pending uncertainty (earnings, litigation, macro). Selling it harvests the premium most of the time but concentrates losses into the tail; the strategy's return profile is short-skew/short-gamma, with negative skew and occasional large drawdowns (2018 "Volmageddon," March 2020). There is, to this author's knowledge, no peer-reviewed study establishing that IV Rank thresholds specifically (50 vs 70) optimize risk-adjusted returns; those numbers are practitioner folklore.

Strengths & limitations

Strengths: both metrics make raw IV comparable and actionable, are trivially computable, and align with a genuine structural edge (VRP) at the index level. IV Percentile is the more statistically defensible of the two because it uses the full distribution rather than two extreme points.

Limitations and the #1 misuse: treating a high reading as a buy/sell signal in isolation, ignoring why IV is elevated. A 90 IV Rank into a binary FDA event is not a free premium harvest — it is correctly priced fear. Other failure modes: IV Rank's extreme sensitivity to a single 52-week spike (it can read "low" while vol is objectively elevated, and vice versa); the arbitrary one-year window (a vol regime shift makes the prior year's range stale); and the fact that neither metric says anything about direction or about whether realized vol will exceed implied this particular time.

Sources

Disputes / flags: The 50/70 "sell premium" thresholds are practitioner heuristics, not empirically optimized — flagged as folklore. The variance risk premium is robust at the index level (Carr-Wu) but weak/contested on single names, where these metrics are most used; this gap is the most important honest caveat in the topic.