How Options Data Providers Calculate the Greeks: Models, Inputs, and Why Platforms Disagree

Jovana A.

— 7 min read
Share
How Options Data Providers Calculate the Greeks: Models, Inputs, and Why Platforms Disagree

There is no official Greek value for any option contract. Delta, gamma, theta, vega, and rho are not facts reported by an exchange — they are model outputs, computed independently by whichever broker, data vendor, or analytics platform is sitting between the raw market data and the number on your screen. Two platforms showing different Greeks for the same contract at the same timestamp are not necessarily in error; they are frequently running different volatility estimates, different pricing models, or different conventions on the same underlying inputs.

This article covers the data layer (what exchanges actually transmit), the standard model (Black-Scholes-Merton, with its closed-form Greeks), the input that introduces most of the variance (volatility, and the implied-volatility solve specifically), and how production systems diverge from the textbook model in practice.

The data layer: exchanges transmit prices, not Greeks

In the US, every listed option trade and quote is consolidated through the Options Price Reporting Authority (OPRA), the SIP (Securities Information Processor) for the options market, administered under an SEC-approved national market system plan. OPRA aggregates last-sale and quotation data from the options exchanges currently participating in the plan — Cboe (BZX, C2, EDGX, Cboe Options), Nasdaq (BX, GEMX, ISE, MRX, PHLX, Nasdaq Options Market), NYSE (American, Arca), MIAX (Options, Pearl, Emerald), BOX, and MEMX — into a single consolidated feed.

The content of that feed is defined narrowly: last-sale reports, the National Best Bid and Offer, contract volume, open interest, and end-of-day summaries. Implied volatility and the Greeks are not part of the core OPRA data product. This is confirmed directly by Cboe's own data shop, which sells "Calcs" (implied volatility and Greeks) as an explicitly optional add-on layered on top of its raw quote and trade history — i.e., even the exchange that administers OPRA treats Greeks as a separate, derived analytics product rather than core market data.

Every downstream Greek you see is therefore computed by whoever owns that layer: your broker's risk engine, a charting platform, or a third-party vendor — each applying its own model, its own volatility estimate, and its own conventions to the same underlying OPRA-sourced quotes.

The standard model: Black-Scholes-Merton

Most production Greek calculations still trace back to the model published by Fischer Black and Myron Scholes in "The Pricing of Options and Corporate Liabilities," Journal of Political Economy, 1973. Robert Merton extended the same framework shortly after to incorporate a continuous dividend yield, which is why the model is frequently denoted Black-Scholes-Merton.

The model is a closed-form solution for the price of a European-style option as a function of six inputs: underlying price (S), strike price (K), volatility (σ), risk-free rate (r), dividend yield (q), and time to expiration (t). With a closed-form price, each Greek is simply a partial derivative of that price with respect to one input, holding the others fixed:

  • Delta (Δ) = ∂price/∂S — sensitivity to a $1 move in the underlying. For a call, Δ = N(d₁), where N is the standard normal CDF.
  • Gamma (Γ) = ∂Δ/∂S — the rate of change of delta itself. Identical formula for calls and puts.
  • Vega = ∂price/∂σ — sensitivity to a one-point change in volatility. Identical for calls and puts.
  • Theta (Θ) = ∂price/∂t — sensitivity to one day's passage of time, holding all else fixed.
  • Rho (ρ) = ∂price/∂r — sensitivity to a change in the risk-free rate.

All five are built from two intermediate terms, d₁ and d₂, themselves functions of log-moneyness (ln(S/K)), volatility, and time to expiration. The full derivations are standard and widely published; the operative point is that every Greek is a model-derived sensitivity, not an observed market quantity.

The model's assumptions are also its known limitations: constant volatility, lognormal returns, and (in the base Black-Scholes form) no early exercise. Real markets satisfy none of these exactly, which is part of why production systems modify or replace the base model — covered below.

The variable input: volatility

Five of the six Black-Scholes-Merton inputs are observable and identical across any competent platform: underlying price, strike, time to expiration, the risk-free rate, and the dividend yield are lookups, not estimates. Volatility is the exception, and it is the primary source of divergence in Greeks between platforms.

There are two distinct ways to populate σ:

Historical (realized) volatility is computed directly from the underlying's past log returns — the standard estimator is the annualized sample standard deviation of daily log returns, σ = √(252) × stdev(ln(Sᵢ/Sᵢ₋₁)). It is objective and fully reproducible given the same price series and lookback window, but it is backward-looking by construction.

Implied volatility (IV) is the model run in reverse: instead of σ → price, the solve is price → σ. Given an observed option price, σ is the value that, substituted into Black-Scholes, reproduces that price. Because the formula has no closed-form algebraic inverse for σ, this is solved numerically — Newton-Raphson iteration is the standard method, with C(S,K,T) (or P(S,K,T)) as the target and the formula's monotonicity in σ guaranteeing a unique root.

The IV solve introduces several provider-specific choices that produce genuinely different — not erroneous — answers for the same contract:

  • Which price is the solve target. Bid, ask, mid, or last trade each produce a different implied volatility for the same contract at the same instant, since the spread itself maps to a range of IVs.
  • Convergence criteria. Starting guess, tolerance, and iteration cap for the root-finder affect the converged value, particularly near the boundaries of the model's validity.
  • Deep out-of-the-money handling. When a contract's bid is quoted (or rounds to) $0.00, solving for IV from that price produces a value that is mathematically valid but economically meaningless — the IV spikes well above what the market is actually pricing. dxFeed has documented its own correction for this: when it detects two consecutive $0.00-bid strikes in a chain, it carries forward the last reliable IV from the preceding liquid strike rather than solving directly on the degenerate quote. This is a vendor-specific patch, not a standard, and it is exactly the kind of invisible methodological choice that produces divergent Greeks across platforms on the same contract.

Model divergence in production systems

Black-Scholes-Merton is not the only model used in production, and the choice matters most for American-style equity options — the instrument type most retail traders actually hold — because the base model is built for European exercise and does not account for early exercise or discrete dividend payments.

Binomial models address this directly by representing the option's life as a discrete-time tree and explicitly valuing the early-exercise decision at each node. They are the standard alternative where early exercise is material. Interactive Brokers' TWS API, for example, returns live Greeks computed by its own internal option-pricing model whenever a client subscribes to market data on a contract — and as with any broker, that calculation is internal and not redistributed unchanged from any exchange.

Vendor-specific adjustments layer further deviation on top of the base model. dxFeed computes Greeks and IV analytically from the standard Black-Scholes formulas, but constructs its own "model-free" price and implied-rate inputs before running the model, rather than taking raw quote data at face value.

Stochastic volatility models, principally the Heston model, relax the constant-volatility assumption directly by letting σ follow its own stochastic process, producing more realistic dynamics — particularly the volatility smile — at the cost of significantly higher computational overhead and reduced transparency versus a closed-form solution. Heston is common in institutional risk and derivatives pricing systems; it is rare in retail-facing platforms, but it marks the upper end of the sophistication spectrum between what a retail broker displays and what a market maker is running internally.

Practical implications

A divergent Greek across two platforms for the same contract is not, by itself, evidence of an error in either system — it is evidence of a different volatility input or a different model, both of which are legitimate and well-documented choices. The practical consequences:

Position sizing and hedge ratios derived from delta should be sourced from a single platform consistently within a strategy, rather than mixed across sources, since the underlying volatility methodology is what's actually driving the discrepancy, not data quality.

Greeks are least reliable — and most likely to diverge sharply between providers — for deep ITM/OTM contracts and near expiration, where the constant-volatility and continuous-pricing assumptions of the base model are weakest and where degenerate (near-zero) quotes can produce unstable IV solves.

When a platform discloses its model (Black-Scholes-Merton vs. binomial vs. other) and its IV solve convention (which price — bid, ask, mid, last — it targets), that disclosure is informative and worth checking before relying on the output for risk management.

For backtesting specifically: if the historical Greeks used in a backtest are computed with a different model or IV methodology than what a live broker uses at execution time, the backtest's reported risk profile and the live position's actual risk profile will diverge — independent of whether the underlying price data itself is accurate.

Sources

  1. Options Price Reporting Authority — Plan Overview. https://www.opraplan.com/
  2. Black, F. & Scholes, M. (1973). "The Pricing of Options and Corporate Liabilities." Journal of Political Economy, 81(3), 637–654.
  3. Cboe Options Institute — "Volatility & the Greeks." https://www.optionseducation.org/advancedconcepts/volatility-the-greeks
  4. Cboe DataShop — "All U.S. Options Exchanges Trade and Quote Data (OPRA Reported)." https://datashop.cboe.com/options-price-bulk-data
  5. Market Data — "Understanding Differences in Implied Volatility and Greeks Between Broker Platforms." https://www.marketdata.app/education/options/differences-in-iv-greeks/
  6. dxFeed — "Greeks and Implied Volatility Provided by dxFeed Market Data Vendor." https://dxfeed.com/data-analytics/greeks-and-implied-volatility/
  7. dxFeed — "dxFeed Updates Implied Volatility and Greeks Calculation." https://dxfeed.com/implied-volatility-and-greeks-calculation-update/
  8. Interactive Brokers — "TWS API v9.72+: Option Greeks." https://interactivebrokers.github.io/tws-api/option_computations.html
  9. Exegy — "What Is Options Price Reporting Authority (OPRA)?" https://www.exegy.com/what-is-options-price-reporting-authority/
  10. Databento Microstructure Guide — "What is the Options Price Reporting Authority (OPRA)?" https://databento.com/microstructure/opra
  11. Macroption — "Black-Scholes Formulas (d1, d2, Call Price, Put Price, Greeks)." https://www.macroption.com/black-scholes-formula/
  12. Wikipedia — "Black–Scholes model." https://en.wikipedia.org/wiki/Black%E2%80%93Scholes_model