The Option Volatility Surface Explained

Executive Summary

The volatility surface is the foundational architectural construct of modern quantitative finance. It serves as the critical interpolation and arbitrage-control layer that connects listed, liquid option market quotes to the advanced mathematical models utilized for derivative pricing, portfolio hedging, and scenario analysis.¹ An implied-volatility surface is not a literal, econometric forecast of future realized volatility. Rather, it is fundamentally a risk-neutral pricing representation—a quoting mechanism that maps option prices into a standardized, dimensionless domain.

Real-world option quotes are characterized by discrete strikes, scattered maturities, illiquidity, and bid-ask noise.¹ Before an options pricing model can be trusted to value complex exotic derivatives or calculate stable risk-management sensitivities (Greeks), it must be strictly calibrated to a continuous, mathematically smooth, and rigorously arbitrage-free surface.¹

This research report investigates the entire life-cycle of the volatility surface. The analysis delineates the transformation of raw market prices into implied volatilities, explores the geometric and economic properties of the volatility smile and term structure, and comprehensively details surface construction methodologies. Furthermore, the report provides a meticulous examination of static no-arbitrage conditions, followed by an in-depth comparison of leading calibration models.

This includes foundational and active research associated with Fischer Black, Myron Scholes, Robert Merton, Bruno Dupire, Jim Gatheral, Antoine Jacquier, Patrick Hagan, and Steven Heston, spanning Stochastic Volatility Inspired (SVI), Surface SVI (SSVI), SABR, Local Volatility (LV), Heston, Rough Volatility, and emerging Machine Learning (ML) frameworks. Finally, the research details practical numerical optimization techniques, dynamic surface hedging assumptions, trading applications, and the model governance frameworks required for production-grade institutional environments.

From Option Prices to Implied Volatility

The Inversion Mechanism of the Black-Scholes Framework

In classical financial mathematics, the foundational models proposed by Fischer Black, Myron Scholes, and Robert Merton assume that the underlying asset price follows a geometric Brownian motion with a strictly constant volatility parameter over time.¹ For futures and forwards, the Black-76 model assumes an identical constant-volatility diffusion process.¹ 

However, market realities—specifically the presence of fat-tailed asset return distributions, asymmetric crash risks, and discontinuous price jumps—dictate that a single constant volatility cannot accurately reproduce observable market option prices.¹ Consequently, the Black-Scholes and Black-76 models are no longer utilized by practitioners as literal descriptions of market dynamics. Instead, they serve as an inversion mechanism—a highly standardized, universally accepted mathematical translator.¹

Implied volatility is defined mathematically as the sole unobservable parameter that, when iteratively input into the Black-Scholes pricing formula alongside observable market inputs, returns a theoretical option value exactly equal to the prevailing executable market price of that option.¹ Traders and automated market-making algorithms conceptualize, communicate, and quote option positions almost exclusively in terms of this implied volatility rather than raw currency premiums.

This is because implied volatility normalizes the mechanical, deterministic effects of the underlying spot price, the strike distance, and the time to maturity, allowing for a pure view of relative option value.¹

Inputs and Practical Data Challenges

To calculate a precise implied volatility, a robust options analytics engine requires an exact specification of several interconnected inputs: the underlying spot or forward price (), the option's strike price (), the exact time to maturity (), the risk-free interest rate curve (), the continuous dividend yield or cost of carry (), the specific option type (Call or Put), and the market price ().¹

Solving for implied volatility () requires a numerical root-finding algorithm, such as the Brent-Dekker method or Newton-Raphson iteration, to equate the market price to the model price: .¹ However, extracting a reliable implied volatility from real-world, high-frequency market data introduces severe practical difficulties:

1. Illiquidity and Wide Bid-Ask Spreads: Deep out-of-the-money (OTM) and deep in-the-money (ITM) options frequently suffer from exceptionally wide bid-ask spreads, making the true "fair" mid-price ambiguous and highly subjective.¹

2. Stale and Crossed Quotes: In high-frequency electronic trading environments, or during volatile macroeconomic data releases, market data feeds may display stale quotes on one side of the limit order book. This leads to crossed spreads or economically meaningless zero-bid quotes that cause root-finding algorithms to fail or return infinite volatilities.¹

3. Discrete Price Ticks: Options trade in discrete minimum price increments (ticks). For low-premium, deep OTM options, a single one-tick movement in the quoted price can mathematically result in a massive, discontinuous jump in implied volatility, introducing artificial, purely mechanical noise into the calibration process.¹

Because of these microstructural frictions, using unadjusted mid-prices alone is demonstrably insufficient. The mid-price of an illiquid, wide-spread option contains virtually no reliable economic information.¹ Consequently, implied volatility must be derived using bid-ask-aware smoothing techniques, computing implied volatilities for the bid and the ask independently to establish a reliable spread-bound, rather than taking a noisy mid-price at face value.¹

Smile, Skew, Term Structure, and the Volatility Surface

Defining the Volatility Smile and Skew

If the Black-Scholes assumption of lognormal asset returns were empirically correct, plotting the implied volatility against the strike price for a fixed maturity slice would result in a perfectly flat, horizontal line.¹ In reality, plotting these variables yields a distinct curve. When this curve slopes upwards on both the ITM and OTM extremes, it is known as a "volatility smile".¹ 

The smile indicates that the market expects fatter tails on both ends of the return distribution than a normal distribution allows. This shape is heavily prevalent in foreign exchange (FX) options and short-dated individual equity options.¹ Conversely, when the curve slopes upward predominantly on the downside (lower strikes) and flattens or declines on the upside (higher strikes), it is termed a "volatility skew".¹ 

The volatility skew became a permanent, structural fixture of global equity index markets immediately following the Black Monday stock market crash of October 1987.¹ Post-crash, market participants recognized that downside strikes required structurally higher implied volatilities to compensate for the severe, asymmetric risk of sudden equity market drawdowns.¹

The Three-Dimensional Surface and Coordinate Systems

Because implied volatility varies simultaneously across both the strike dimension and the maturity dimension, these two axes combine to form a three-dimensional implied-volatility surface mapping .¹ 

The strike dimension captures asymmetric crash risk, tail demand, and kurtosis, while the maturity dimension captures the term structure, encoding event risk, mean reversion, and the aggregate uncertainty scaled over the investment horizon.¹

To build robust mathematical models, quantitative developers must transform raw strikes into normalized, dimensionless coordinate systems. The choice of coordinate system fundamentally dictates the numerical stability of the calibration.

Coordinate System	Practitioner Application and Assessment
Raw Strike	Intuitive for retail traders but mathematically problematic because it does not scale with the underlying forward price or time to maturity.1
Delta	Standard convention in FX options markets. It standardizes moneyness by the approximate probability of the option expiring ITM, naturally adjusting for time and volatility.5
Log-Moneyness	Centers the At-The-Money (ATM) point at zero, but fails to account for the drift caused by interest rates and dividends over time.1
Log-Forward Moneyness	Highly preferred for equity surfaces. It removes the deterministic drift of interest rates and dividends, perfectly centering the ATM forward point at zero across all maturities.1
Total Implied Variance	The optimal, critical coordinate system for mathematical surface calibration. It elegantly separates strike, carry, and time, simplifying the enforcement of strict no-arbitrage bounds.1

Economic Interpretation of the Surface Shape

The topological shape of the volatility surface serves as an economic fingerprint of the market's risk-neutral probability distribution. Foundational work by Breeden and Litzenberger demonstrated that by taking the second derivative of call option prices with respect to the strike price, practitioners can extract the exact market-implied risk-neutral probability density function.

In equity index options, the persistent downside skew is fundamentally driven by the structural demand for crash protection.¹ Institutional investors natively hold long equity portfolios and systematically purchase OTM puts to hedge against drawdowns, driving up the prices—and thus the implied volatilities—of those lower strikes. Conversely, upside call overwriting (yield enhancement strategies) structurally suppresses the implied volatilities of higher strikes.

This skew is also deeply intertwined with the "leverage effect." As equity prices fall, corporate debt-to-equity ratios mathematically rise, rendering the underlying companies structurally riskier and thus legitimately more volatile.⁷ Furthermore, jump risk contributes heavily to the steepness of short-dated skew. Because continuous diffusion processes (like standard Brownian motion) mathematically cannot produce massive price moves over infinitesimally short time horizons, the steep skew in short-dated options mathematically implies that the market is pricing in the probability of discontinuous, instantaneous price jumps, as explored extensively by Nassim Taleb and Robert Merton.¹

The surface term structure is heavily impacted by scheduled events. Short-dated surfaces frequently display pronounced "kinks" or isolated bumps around scheduled macroeconomic data releases, corporate earnings announcements, central-bank interest rate decisions, and national elections.⁴ This event variance is priced directly into the specific expiration encompassing the date. Once the scheduled event passes, the localized event variance instantly drops out of the term structure, and the surface globally smooths out.⁴

Surface Construction and Data Cleaning

Constructing a stable, production-grade volatility surface requires a highly systematic, rigid pipeline to transition from noisy raw market data to a mathematically sound pricing object.¹

The complete workflow involves several sequential stages:

1. Data Ingestion and Aggregation: The system obtains real-time or end-of-day option-chain data, capturing bid, ask, and traded volume across all available strikes and expirations.

2. Estimation of Forwards and Discount Factors: Reliable forward modeling is the most critical precursor to surface construction. Using put-call parity on the most liquid ATM strikes is standard practice to back out implied forward prices and synthetic continuous dividend yields, ensuring that the call and put surfaces align flawlessly without embedded arbitrage.¹

3. Filtering and Cleaning Quotes: Raw quotes must be aggressively filtered. Stale quotes, crossed bid-ask spreads, and options with zero bid volume are systematically removed.¹

4. Inversion to Implied Volatility: Filtered prices are inverted using the Black-Scholes or Black-76 formulas to extract bid-implied and ask-implied volatilities.

5. Fitting the Maturity Smiles: Individual maturity smiles are fitted using parameterized models, enforcing convexity constraints.

6. Interpolation and Extrapolation: The algorithm interpolates the total variance smoothly between the fitted maturity slices. Finally, the surface must extrapolate reasonably into the extreme, unobservable wings.

7. Arbitrage Testing and Validation: The interpolated surface is subjected to rigid derivative tests to ensure the absence of static arbitrage, followed by validation to confirm all model prices fall within observed market bid-ask bounds.¹

The Treatment of Extreme Strikes and Bid-Ask Weighting

Discarding deep ITM and deep OTM wings completely during the filtering stage is a dangerous practice, as it blinds the calibration model to critical tail-risk information embedded in the market. Instead, best practice dictates that these "noisy wings" are retained but heavily downweighted during the numerical optimization.¹

Giving every option quote equal weight inevitably causes the calibration to overfit to market noise in illiquid strikes.¹ Institutions like FactSet utilize bid-ask-aware and vega-weighted calibration frameworks.¹ Vega weighting mathematically forces the optimization algorithm to heavily anchor the model to the reliable, highly liquid ATM options (which possess the highest vega sensitivity and the tightest bid-ask spreads), while allowing the model to smoothly deviate from the noisy mid-prices in the extreme, low-vega wings.¹

Furthermore, smoothing directly in implied volatility space is a known anti-pattern that destroys convexity.¹ Advanced methodologies, such as Fengler's method, apply arbitrage-free constrained splines directly in option-price space, smoothing the prices first under strict shape constraints before inverting them back to implied volatilities.¹

No-Arbitrage Conditions

A visually smooth, aesthetically pleasing volatility surface may still contain catastrophic hidden arbitrage.¹ If a pricing model is calibrated to a surface containing static arbitrage, the model will output negative probabilities, leading to the severe mispricing of exotic derivatives and the generation of explosive, unstable risk sensitivities. There are two paramount static arbitrage conditions that must be rigidly avoided:

1. Butterfly Arbitrage and Negative Probability Densities

Butterfly arbitrage occurs when a specific combination of options (e.g., a long butterfly spread consisting of buying a low strike, selling two middle strikes, and buying a high strike) has a negative theoretical price. Economically, this implies an impossible negative probability of the underlying asset expiring exactly at that middle strike.¹

2. Calendar Spread Arbitrage

Calendar spread arbitrage materializes if a longer-dated option is priced cheaper than a shorter-dated option with identical strikes. Because options inherently possess non-negative time value (assuming positive interest rates and no extreme dividend effects), the total variance must strictly increase with time. Assuming proportional continuous dividends, the surface is free of calendar spread arbitrage if and only if the total variance  is monotonically non-decreasing in  at every fixed log-strike :

If this strict condition is violated, total variance lines cross on a plot, permitting traders to extract risk-free profit by executing calendar spreads.¹ Total implied variance is heavily preferred for interpolation precisely because calendar arbitrage is trivial to mathematically detect and enforce in this space—the lines simply cannot cross.¹

Furthermore, inconsistent wing extrapolation can violate Roger Lee’s moment formulas. Lee's theorems prove that implied variance cannot grow faster than linearly in the extreme wings (). If a model extrapolates variance quadratically, it implies an infinite option price, allowing for asymptotic arbitrage.¹

Local, Stochastic, and Rough Volatility Dynamics

It is critical to distinguish between an implied-volatility surface (which is purely a static quoting mechanism) and the actual model dynamics used for pricing and hedging.¹

Dupire Local Volatility

Local Volatility (LV), pioneered by Bruno Dupire, Emanuel Derman, and Iraj Kani, assumes that the instantaneous volatility of the underlying asset is a deterministic function of both time and the asset's current price: .¹ The LV surface is mathematically extracted directly from the implied volatility surface. Lorenzo Bergomi and Jim Gatheral formulated Dupire's equation cleanly in terms of total variance  and log-moneyness :

This powerful formula demonstrates that local volatility is heavily dependent on the partial derivatives (both the slope and the curvature) of the implied variance surface.¹

Stochastic Volatility (Heston)

Stochastic Volatility (SV) models, such as the widely adopted Heston model, introduce a separate stochastic differential equation (SDE) for the variance process itself, incorporating a mean-reversion speed, a long-term variance level, and a "volatility of volatility" (vol-of-vol) parameter.¹

• The Trade-off: Because SV generates realistic, persisting future smiles that move dynamically with the spot price, it is vastly superior for delta and vega hedging over time. However, standard continuous SV models fundamentally lack the mathematical capability to fit the extreme steepness of short-dated equity skew.¹ To resolve this, Stochastic-Local-Volatility (SLV) models combine an SV backbone with a deterministic LV "leverage function" to force an exact fit, though calibration becomes exceptionally computationally expensive.¹

Rough Volatility

Empirical evidence compiled by Peter Friz, Jim Gatheral, and others indicates that historical volatility time series exhibit scaling behavior consistent with Fractional Brownian Motion (fBm), where the Hurst exponent () is significantly less than the standard  (typically empirically measured at ).⁷ This "roughness" accounts for the explosive steepness of the short-term implied volatility skew.¹⁶ 

Rough volatility models possess the unique capability to jointly calibrate to both SPX and VIX options simultaneously—a historically impossible feat for standard diffusion models.¹⁶ Calibration involves mapping the decay speed of the implied volatility across strikes directly to the Hurst parameter, though the non-Markovian nature of fBm requires intense Monte Carlo simulation, complicating real-time pricing.⁷

Surface Dynamics and Hedging Assumptions

A perfectly calibrated surface provides a static snapshot of current market prices. However, active risk management requires assumptions about how the surface will physically deform when the underlying spot price moves over the next hour or day. Practitioners utilize heuristic "sticky" rules of thumb.⁴

1. Sticky-Strike: Assumes that the implied volatility assigned to an absolute strike  remains constant regardless of spot movement. This dynamic is commonly applied in single-name equity options where specific strike levels act as psychological or technical barriers.²¹

2. Sticky-Delta (Sticky-Moneyness): Assumes that the implied volatility at a given delta or moneyness metric remains constant. As the spot price falls, the entire volatility skew physically shifts downwards with it. This is the absolute standard assumption in FX markets and broad equity index markets.⁵

3. Sticky-Implied Tree: An assumption pioneered by Emanuel Derman, intimately tied to the deterministic dynamics implied by local volatility models, where local volatility at a specific spot and time remains invariant.²¹

Traders must compute "smile-adjusted" Greeks to hedge effectively. A standard Black-Scholes Delta assumes volatility is constant as spot moves. However, under a sticky-delta regime, a "shadow delta" (or smile-adjusted delta) must factor in the explicit slope of the skew.⁴ It mathematically recognizes that a drop in the spot price will concurrently trigger an expansion in the implied volatility (due to the skew), actively altering the required hedge ratio.⁴

Trading and Risk-Management Applications

The calibrated implied volatility surface powers multiple sophisticated trading infrastructures and risk methodologies:

• Higher-Order Greeks and FX Pricing: In FX options markets, exotic derivatives like barrier options (e.g., Double-No-Touch) are priced and hedged using the Vanna-Volga pricing method, also known as the Traders' Rule of Thumb.²² Vanna () measures the sensitivity of Vega to spot moves, explicitly compensating for the skew (Risk Reversals). Volga or Vomma () measures the convexity of Vega, compensating for the curvature of the smile (Butterflies).²² The Vanna-Volga method calculates adjustment costs using a replicating matrix of At-The-Money straddles, 25-delta Risk Reversals, and Vega-weighted Butterflies, adding this cost to the base Black-Scholes price.²²

• Volatility Arbitrage and Relative Value: Quantitative trading desks monitor the surface continuously to identify localized mispricings, such as depressed wings or disjointed term structures. They deploy delta-neutral calendar spreads or butterfly spreads to capture the mean-reversion to historical norm parameters.⁴

• Dispersion Trading: Traders map the implied volatility surface of a broad index (like the S&P 500) against the vega-weighted sum of the individual constituent volatility surfaces. This strategy structurally isolates and trades implied correlation, capturing the correlation risk premium.²⁷

• Scenario Stress Testing: Enterprise risk engines systematically perturb the  (level) and  (skew) SVI parameters across the entire global portfolio to simulate historic market crashes (e.g., 2008, 2020), ensuring institutional capital survival under severe tail-risk materialization.²⁸

Model Risk and Governance Considerations

A theoretically brilliant mathematical model is practically useless—and potentially catastrophic—without stringent model governance. Driven largely by post-2008 regulatory frameworks (such as SR 11-7 and OCC 2011-12), Model Risk Management (MRM) enforces a rigid "three lines of defense" hierarchy within financial institutions.³

1. First Line (Front Office / Quants): The model owners who develop the calibration framework. They are responsible for establishing automated data-quality controls, continuous calibration monitoring, and programming explicit fallback methods (e.g., automatically falling back to yesterday's validated surface if today's SVI optimization fails to converge due to a market dislocation).²⁸

2. Second Line (Independent Model Validation): A mathematically proficient, fully independent risk team that backtests the calibration. They stress-test the model to ensure the optimizer does not overfit to noise, verify that parameter stability is maintained over time, and confirm that the model behaves rationally during simulated liquidity crises.²⁹ They hold the authority to veto the model's deployment.²⁸

3. Third Line (Internal/External Audit): Evaluates the overarching governance architecture, ensuring that strict version control, extensive audit trails, reproducibility standards, and regulatory compliance are intact.²⁸

A critical component of modern MRM is verifying that the selected model strictly suits its intended purpose. A polynomial spline model might pass an in-sample pricing error test perfectly but fail spectacularly under stress testing due to extreme extrapolation instability. This highlights the absolute necessity of deep qualitative validation extending far beyond simple numerical error metrics.³

Conclusion

A robust, production-grade volatility-surface calibration framework is a delicate, highly engineered synthesis of empirical market microstructure, advanced stochastic calculus, and heavily constrained numerical optimization.

The most critical realization for practitioners is that the model which best fits the raw market quotes is rarely the best model for pricing, hedging, or risk management.¹ Models that perfectly fit raw data, such as unconstrained splines, inevitably embed bid-ask noise into their derivatives, triggering negative probability densities, broken local volatility surfaces, and highly unstable hedge ratios.¹ 

Conversely, heavily constrained parametric frameworks like SSVI intentionally sacrifice millimeter-perfect in-sample fits to mathematically guarantee global arbitrage-freeness, asymptotic stability, and robust, reliable Greeks.¹

The transition from a static implied volatility surface to a dynamic, forward-looking risk-management environment requires carefully navigating the fundamental trade-offs between exact, static calibration (Dupire Local Volatility) and realistic, dynamic future paths (Heston Stochastic Volatility and Rough Volatility).¹ 

Ultimately, surface calibration is an ill-posed inverse problem; resolving it requires strict algorithmic discipline, bid-ask awareness, and rigid institutional model governance.

Works cited

1. The Volatility Surface: Option Calibration and Construction Pipeline

2. modelling and implementation of local volatility surfaces, accessed July 24, 2026, http://etd.lib.metu.edu.tr/upload/12617955/index.pdf

3. The Evolution of Model Validation: Importance of Governance - LSEG, accessed July 24, 2026, https://www.lseg.com/en/post-trade/solutions/podcast/ahead-of-the-curve/evolution-of-model-validation-governance-importance

4. Volatility Smile: What It Is & Why It Matters in 2026 - Quantt, accessed July 24, 2026, https://www.quantt.co.uk/resources/volatility-smile-explained

5. Volatility Smile and Delta Hedging (Part 2) — Down the rabbit hole of Smile Risk Hedging. - Harel Jacobson - Medium, accessed July 24, 2026, https://volquant.medium.com/volatility-smile-and-delta-hedging-part-2-down-the-rabbit-hole-of-smile-risk-hedging-7867e955bb3b

6. No Arbitrage SVI - SIAM Publications Library, accessed July 24, 2026, https://epubs.siam.org/doi/pdf/10.1137/20M1351060?download=true

7. American Option Pricing Under Time-Varying Rough Volatility: A Signature-Based Hybrid Framework - arXiv, accessed July 24, 2026, https://arxiv.org/html/2508.07151

8. Proof of Durrleman's condition for the implied volatility surface to eliminate butterfly spread arbitrage - Quantitative Finance Stack Exchange, accessed July 24, 2026, https://quant.stackexchange.com/questions/82286/proof-of-durrlemans-condition-for-the-implied-volatility-surface-to-eliminate-b

9. Deep Smoothing of the Implied Volatility Surface, accessed July 24, 2026, https://papers.neurips.cc/paper_files/paper/2020/file/858e47701162578e5e627cd93ab0938a-Paper.pdf

10. The SVI implied volatility model and its calibration - DiVA Portal, accessed July 24, 2026, https://www.diva-portal.org/smash/get/diva2:744907/FULLTEXT02.pdf

11. SABR volatility model - Wikipedia, accessed July 24, 2026, https://en.wikipedia.org/wiki/SABR_volatility_model

12. Interest rate volatility - II. SABR and its flavors - Baruch MFE Program, accessed July 24, 2026, https://mfe.baruch.cuny.edu/wp-content/uploads/2015/06/VolWork2-Andrew.pdf

13. The SABR Model: Modelling the Volatility Smile - CBS Research Portal, accessed July 24, 2026, https://research.cbs.dk/files/98729776/1617240_modelling_the_volatility_smile_v20230514.pdf

14. HJM Local Volatility Model. arXiv:2301.13595v8 [q-fin.PR] 30 Jan 2025, accessed July 24, 2026, https://arxiv.org/pdf/2301.13595

15. Dupire Local Volatility with Cash Dividends Part 2 - Chase the Devil, accessed July 24, 2026, https://chasethedevil.github.io/post/dupire_cash_dividend_part2/

16. A SURVEY OF ROUGH VOLATILITY | International Journal of Theoretical and Applied Finance - World Scientific Publishing, accessed July 24, 2026, https://www.worldscientific.com/doi/10.1142/S0219024925300021

17. Rough volatility: An overview - Baruch MFE Program, accessed July 24, 2026, https://mfe.baruch.cuny.edu/wp-content/uploads/2017/05/RoughVolatilityBarcelona2017.pdf

18. Trading Volatility Roughness — Rethinking Statistical Arbitrage - Harel Jacobson - Medium, accessed July 24, 2026, https://volquant.medium.com/trading-volatility-roughness-rethinking-statistical-arbitrage-41a204c93b6d

19. No arbitrage SVI - arXiv, accessed July 24, 2026, https://arxiv.org/pdf/2005.03340

20. Volatility Surfaces: Theory, Rules of Thumb, and Empirical Evidence - University of Toronto, accessed July 24, 2026, https://www-2.rotman.utoronto.ca/~hull/DownloadablePublications/DHSPaperdraft7.pdf

21. Lecture 9: Patterns of Volatility Change - Copyright Emanuel Derman 2008, accessed July 24, 2026, https://emanuelderman.com/wp-content/uploads/2013/09/smile-lecture9.pdf

22. Detailed Explanation of the Vanna-Volga Pricing Method - Mathema Documents, accessed July 24, 2026, https://help.mathema.com.cn/latest/docs/pricing/vv

23. Vanna-volga pricing - EconStor, accessed July 24, 2026, https://www.econstor.eu/bitstream/10419/40192/1/573774811.pdf

24. Vanna–Volga pricing - Wikipedia, accessed July 24, 2026, https://en.wikipedia.org/wiki/Vanna%E2%80%93Volga_pricing

25. The vanna - volga method for derivatives pricing. - Munich Personal RePEc Archive, accessed July 24, 2026, https://mpra.ub.uni-muenchen.de/36127/1/MPRA_paper_36127.pdf

26. Vanna Volga and Smile-consistent Implied Volatility Surface of Equity Index Option Kun Huang - ACFR, accessed July 24, 2026, https://acfr.aut.ac.nz/__data/assets/pdf_file/0017/185300/172496-K-Huang-Vanna-Volga_Auckland.pdf

27. Volatility - Core Concepts - NY Institute of Finance, accessed July 24, 2026, https://www.nyif.com/volatility-core-concepts.html

28. Tech-driven model risk management – Driving value and efficiency, accessed July 24, 2026, https://www.risk.net/risk-management/6764106/tech-driven-model-risk-management-driving-value-and-efficiency

29. Model risk management - KPMG International, accessed July 24, 2026, https://assets.kpmg.com/content/dam/kpmg/pdf/2016/06/KPMG-Whitepaper-Model-Risk-Management-2016.pdf

30. Risk Modeling: Definition, Methods, Workflow & Challenges - Galorath, accessed July 24, 2026, https://galorath.com/risk/risk-modeling/