Why implied volatility and Greeks differ between platforms

Pull up the same option on two platforms and the implied volatility is often a point or two apart, sometimes ten on a contract expiring today. The deltas disagree with it. A common first guess is stale data. More often, both platforms computed a correct number from different inputs.

Implied volatility is not in the market data. The options feed carries prices. Every IV and every Greek on every platform was calculated by that platform, and the choices behind it are rarely shown next to the number.

How it works

The consolidated US options feed is OPRA. Its binary specification, version 6.4b of 25 August 2026, defines quotes, trades, open interest and end-of-day summaries. No field in it is an implied volatility or a calculated Greek. The only delta is a negotiated term on certain FLEX trades. There is no underlying stock price on the quote or trade messages either. The spec's only underlying-price field is in the end-of-day summary. The stock price comes from a separate equity feed, on a separate clock.

So every IV is solved backwards. Take a pricing model, fill in the stock price, strike, time to expiry, interest rate and dividends, and search for the volatility at which the model's value equals the option's price. The Greeks are then the model's sensitivities at that volatility. Every input is a choice:

1. The model. A closed-form European formula, or an American-exercise method.
2. The price. Bid, ask, mid or last trade.
3. The stock price. Which feed, and taken at which instant relative to the option quote.
4. The clock. Calendar time or trading time, whole days or minutes.
5. The rate and the dividend. Which published rate, and which dividend forecast.

The implied volatility and greeks glossary entries define the terms. This page is about the size of each choice, with numbers.

The examples below use the textbook Black-Scholes formula and invented quotes, with a 4% rate and no dividend unless stated. They are illustrations computed for this page, not any platform's output.

The model: European formula or early exercise

US single-stock options are American. Cboe's equity options specifications say they "generally may be exercised on any business day up to and including on the expiration date". SPX index options are European. Cboe's SPX specifications say they "generally may be exercised only on the expiration date". Black-Scholes prices the European kind. For an American option it ignores the right to exercise early. That right is worth most on in-the-money puts and on calls just before a dividend.

Libraries offer both, and more than two. QuantLib's vanilla engines include an analytic European engine, binomial trees, finite differences, and approximations for American options such as Barone-Adesi–Whaley and Bjerksund–Stensland. A vendor's documentation shows the choice in practice. ThetaData's Greeks article says it "uses the Black Scholes for all Greeks calculations". It also offers separate binomial endpoints on a Leisen-Reimer tree that "allows early exercise at every node", and solves IV on that tree for puts and for calls with a dividend supplied. Its own warning follows: "On contracts solved on the tree, implied_vol will not match the Black-Scholes endpoints for the same contract."

How much? Take a put struck at 110 on a 100-dollar stock, 60 days to expiry, priced at 10.60. Black-Scholes, treating it as European, gives an IV of 26.8%. A 1,000-step American binomial tree gives 24.9%. Same contract, same price, two points apart, and every Greek moves with it.

The price: bid, ask, mid or last

A quote is two prices, and each gives its own IV. Cboe's volatility index methodology, version 5.0 of 26 February 2026, commits to one. Its option prices "reflect the midpoint of each candidate constituent option series' bid / ask quotes", and series with null quotes or a bid above the ask are excluded. ORATS publishes callMidIv and putMidIv, "mid implied volatility", alongside smvVol, which it calls its "final implied Volatility", a smoothed value. A platform that shows the IV of the last trade is using a price that may be hours old.

The spread sets the size. A call on a 100-dollar stock, struck at 100, quoted 0.90 bid and 1.10 ask on a Friday at 3 p.m. for a Monday 4 p.m. expiry, has an IV of 24.3% at the bid, 27.0% at the mid and 29.8% at the ask, counting time in calendar minutes. On a wide market in a thin contract, the gap between bid IV and ask IV can exceed any other difference on this page.

The clock: how long until expiry

Volatility is quoted per year, and implied volatility scales with the square root of the time input. So the way a platform counts time changes IV in proportion. There are four common choices.

• Calendar minutes. Cboe's methodology measures time "in calendar years", dividing minutes to expiration by 525,600, the minutes in a 365-day year.
• Trading time. Only the hours the market is open count, as a fraction of a 252-day year.
• Whole days. One number of days to expiration, with the intraday fraction dropped.
• A rule for the last day. ThetaData documents that for contracts with fewer than 7 days left, days to expiration come from the quote timestamp, and longer ones use a whole number. Its legacy calculation "applies a fixed DTE of 0.15 on same-day expiration".

The Friday call above, at its 1.00 mid, is 27.0% counting calendar minutes and 36.7% counting trading time. Over the weekend, calendar time runs and trading time does not.

On the expiry day the conventions part completely. The same call with two hours to go, at 0.25 mid, has an IV of 41.4% counting calendar minutes, 17.8% counting trading hours, 30.8% with a fixed 0.15 days and 11.7% if the remaining time is rounded up to one day. Same option, same quote, four numbers between 12% and 41%. This is why IV on options expiring today looks broken from one platform to the next.

The stock price and its timestamp

The underlying price comes from another feed. ThetaData says it uses "the exact underlying tick (price) at the time of the option tick". A platform that pairs an option quote with a stock price cached a second earlier, or with the last trade rather than the midpoint of the stock's quote, solves a different problem. For a 30-day at-the-money call priced at 2.50, a stock price of 100.00 gives an IV of 20.4% and a price of 100.20 gives 19.5%. On an active stock, twenty cents can pass between two quotes a second apart.

Delta moves for the same reason, before any change in IV, because delta is a function of the stock price. The options data guide covers why the two feeds are separate purchases.

The rate and the dividend

A rate is needed to discount the strike, and there is more than one rate to choose. Cboe derives its rate from Treasury constant-maturity yields, interpolated with a cubic spline and converted from bond-equivalent to continuously compounded. ThetaData uses SOFR by default and notes that "SOFR is reported 1 day after the report date", so today's calculation uses yesterday's rate. The New York Fed publishes SOFR each business day at about 8 a.m. Eastern. Even one Treasury bill is quoted two ways on the Treasury's daily bill rates, as a bank discount on a 360-day year and as a coupon equivalent on a 365- or 366-day year. On a short-dated option the rate barely matters. On a one-year at-the-money call priced at 9.00, a rate of 4.0% gives 17.6% and 4.5% gives 16.9%.

Dividends are a forecast, and platforms forecast differently. ThetaData "currently ignores dividends" unless the user supplies one. ORATS uses an annual figure "from the next year of expected dividends" and also publishes a residualRate, defined as "implied interest rate data": a rate read out of option prices rather than taken from a published series. Cboe's methodology avoids the forecast altogether: it infers a forward price from the call and put at the strike where their prices are closest. For a 45-day at-the-money call priced at 3.00, assuming no dividend gives 19.7% and assuming a 3% yield gives 21.1%.

IV rank and IV percentile: same names, different windows

IV rank and IV percentile are statistics on a history of implied volatility, so they inherit every choice above and add two more: which IV series, and which window.

tastylive defines IV rank as where current IV sits between the 52-week low and high, and IV percentile as the share of days in the last 52 weeks with lower IV, computed by dividing that count "by 252 trading days". ORATS publishes ivRank1m and ivRank1y, with the formula "(Current IV - 1 month Low IV) / (1 month Max - 1 month Min)" for the first, and ivPct1m and ivPct1y for percentile. A one-month rank and a one-year rank of the same stock on the same day can be at opposite ends of the scale.

The series underneath differs too. ORATS's iv30d is a "30 calendar day interpolated implied volatility". Another platform may rank the IV of the nearest expiry, or of one at-the-money contract. An IV rank series also inherits every discontinuity in the IV history beneath it. Splicing two providers' histories, or one provider's history across a methodology change, creates a jump the rank will treat as real.

What you can do about it

Ask for the five inputs, and read them from the documentation. Model and exercise style, which price is inverted, the stock price and its timestamp, the time convention, and the rate and dividend source. A platform that documents them can be reconciled. One that does not cannot, and its IV is comparable only with itself.

Reproduce one contract yourself. Take the platform's displayed IV, the stock price it shows, its rate if published, and the strike and expiry. Feed them into one model, for example QuantLib's European engine and then an American one, and see which reproduces the platform's option price. Then change one input at a time. The input that closes the gap is the explanation, usually within a few minutes.

Compare like with like before comparing platforms. Mid IV with mid IV, at the same moment, same time convention. Many apparent disagreements disappear once the bid-ask choice and the calendar-or-trading-time choice match. On contracts expiring within a day, do not compare IVs across platforms at all unless both document their time convention.

Treat IV rank as platform-specific. Write down the window and the IV series next to any IV rank you record. Do not carry a threshold from one platform to another, and do not compute a rank over a history spliced from two providers.

When it has to be consistent, buy the methodology rather than the number. A vendor that publishes its rate, dividend and smoothing choices, as ORATS does in its field definitions, gives you a series you can compare across years. The broader field, from chains to flow, is in options analytics, and how to get an options chain covers what the free tiers return.