Skip to content
VolSurface Desk
Directory

FX Risk Reversal Strangle Surfaces

Malz Quadratic vs Vanna-Volga FX Volatility Smile Interpolation

Graduate from Malz (1997) delta-parabola approximations to institutional Castagna-Mercurio Vanna-Volga pricing: construct exact hedging weights x1, x2, x3 across 25d Put, ATM, and 25d Call to price any FX strike without arbitrage.

By Dr. Alistair Finch, CQF, FRMReview not recorded · Updated
Malz Quadratic vs Vanna-Volga FX Volatility Smile Interpolation — VolSurface Desk 2026 quantitative & supervisory data card
VolSurface Desk (FX 25-Delta Risk Reversals, 10-Delta Butterflies & Event Vol Weights): Hedging-consistent 3-pillar Lagrange weights vs Malz (1997) delta-space parabolas.Source: VolSurface Desk Institutional Research DeskLicense: CC-BY-4.0

When an interbank FX market maker only observes three liquid market quotes on a 2-month tenor (K1 = K25P, K2 = KDNS, K3 = K_25C), how should they price an off-pillar corporate client request at a 14-delta call or 37-delta put? Historically, FX desks progressed from Allan Malz's (1997) delta-parabola to Castagna & Mercurio's (2007) Vanna-Volga replication—a hedging-driven smile engine that still powers tier-1 G10 FX exotic desks today.

Within the FX 25-Delta Risk Reversals, 10-Delta Butterflies & Event Vol Weights silo and our FX Volatility Surfaces & Event Vol Guide, comparing Malz, Vanna-Volga, and SABR/SVI clarifies when three-pillar replication excels and where 10-delta wings require full stochastic local volatility.

Exact Lagrange Weights of Vanna-Volga vs the Malz Delta Formula#

In Malz (1997), volatility is written directly as a parabola in signed call delta Δ ∈ (0, 1): σMalz(Δ) = σATM − 2·RR25·(Δ − 0.5) + 16·BF25·(Δ − 0.5)2 (note this is implicit in K because Δ itself depends on σ_Malz(Δ)!).

By contrast, Vanna-Volga starts from an economic no-arbitrage hedging portfolio: in a stochastic-volatility market where ATM volatility σ2 = σATM fluctuates randomly, we neutralize a target option V(K)'s Vega, Vanna, and Volga/Vomma using weights (x1, x2, x3) in the three liquid benchmark options at strikes K1 < K2 < K3. Solving that 3 × 3 Greek system at σ = σATM yields closed-form Lagrange log-moneyness weights:

Quantitative Pricing Model: Malz Quadratic vs Vanna-Volga FX Smile Interpolation Guide#

Formula or model from the articleNot independently verified
V_VV(K) = V_BS(K, σ_ATM) + Σ_{i ∈ {1, 2, 3}} x_i(K) · [ V_mkt(K_i, σ_i) − V_BS(K_i, σ_ATM) ]
  • Closed-Form Hedging Weights (i = 1,2,3; j,m ≠ i): xi(K) = [ VegaBS(K, σATM) / VegaBS(Ki, σATM) ] × [ ln(K / Kj) · ln(K / Km) ] / [ ln(Ki / Kj) · ln(Ki / Km) ] !
  • Exact Pillar Recovery Check: At K = K1 (25d Put), ln(K1 / K1) makes x2(K1) = 0 and x3(K1) = 0 while x1(K1) = 1 => VVV(K1) ≡ Vmkt(K1) with ZERO error!
  • Worked EUR/USD 2-Month 15-Delta Put Pricing (S = F = 1.0850, K1[25P] = 1.0662 @ 7.95%, K2[ATM] = 1.0858 @ 7.20%, K3[25C] = 1.1049 @ 7.05%, Target K[15P] = 1.0558):
  • Log-Lagrange Ratios: l1(K) = +1.742, l2(K) = −0.896, l3(K) = +0.154 => Vanna-Volga Smile Cost Over BS(7.20%) = +21.8 pips => Inverted σ_VV(15P) = 8.48% vs Malz Parabola = 8.31% (−0.17v wing underpricing by Malz!).

Traders verify individual pillar Black-Scholes Vegas and option prices in our Black-Scholes Option Calculator and invert V_VV(K) back into implied volatility via the Implied Volatility Calculator.

Castagna-Mercurio Vanna-Volga 3-Pillar Pricing & Hedging Engine#

Economic hedge replication pipeline of Vanna-Volga pricing from target strike Greeks to 3-pillar market over-cost
Economic hedge replication pipeline of Vanna-Volga pricing from target strike Greeks to 3-pillar market over-costVerified Exhibit

Notice Metric #1 (Only i = 1 & i = 3 active!): because σ2 = σATM, the middle pillar's market-minus-BS over-cost Vmkt(K2, σATM) − VBS(K2, σATM) is identically zero! Thus Vanna-Volga option price equals flat ATM Black-Scholes plus just **two hedge adjustments (x1 · OverCost25P + x3 · OverCost_25C)**!

Comparing Four FX Smile Interpolation Methodologies#

Quantitative Distribution · Market Input Pillars
Market Input PillarsDeep Wing Behavior (< 5d)
013-Point Vanna-Volga25P, ATM, 25C025-Point SVI / SABR10P, 25P, ATM, …03Heston-SLV PDEFull 5-pillar g…
Smile ModelMarket Input PillarsHedging InterpretationDeep Wing Behavior (< 5d)
Malz (1997) Delta ParabolaATM, RR25, BF25Pure empirical polynomial in ΔUnderstates jump kurtosis; flat beyond Δ → 0, 1
3-Point Vanna-Volga25P, ATM, 25CExact Vega–Vanna–Volga static hedge!Smooth to ~8d; needs wing damping below 5d
5-Point SVI / SABR10P, 25P, ATM, 25C, 10CParametric stochastic-vol asymptoticsObeys Roger Lee linear total-variance wing slope!
Heston-SLV PDEFull 5-pillar grid across all tenorsDynamic spot + stochastic variance PDEGold standard for path-dependent TARFs & PRDCs

10-Delta Wing Extrapolation Error When Calibrated Only to 25-Delta#

Implied volatility fit residual against market 10-delta broker quotes across Malz Vanna-Volga SABR and 5-point SVI
Implied volatility fit residual against market 10-delta broker quotes across Malz Vanna-Volga SABR and 5-point SVIVerified Exhibit

Look at Bar 3 (+0.36v excess kurtosis on EM 10d wings): why do emerging-market and central-bank jump currencies trade with 10-delta butterflies far richer than 25-delta Vanna-Volga predicts? That brings us directly to FX 10-Delta Wing Butterflies & Jump Kurtosis Pricing!

Frequently asked questions

Why did central banks like the Fed and Bank of England adopt the Malz (1997) smile formula?

Central-bank economists needed a simple, closed-form formula to extract risk-neutral exchange-rate probability densities from historical 3-pillar (ATM, 25d RR, 25d BF) broker archives without running iterative stochastic-volatility PDE solvers!

How does Vanna-Volga compute the exact hedging weights (x1, x2, x3) for an arbitrary strike K?

Inside a constant-σ_ATM Black-Scholes world, Vanna and Volga of any option are simple polynomials of log-moneyness ratios! Solving the 3×3 linear system matching Vega, Vanna, and Volga reduces beautifully to Lagrange interpolation polynomials in ln(K/K1), ln(K/K2), and ln(K/K3)!

Where does 3-point (25P, ATM, 25C) Vanna-Volga break down in the deep wings?

Beyond ~5-delta wings, first-order Vanna-Volga price adjustments can grow faster than Roger Lee’s √|k| asymptotic variance bound or violate Durrleman butterfly positivity if 10-delta quotes diverge sharply from 25-delta extrapolation—requiring SSVI or Heston-SLV wing Blending!

What is the difference between First-Order and Second-Order Vanna-Volga implied volatility approximations?

First-order VV divides the Vanna-Volga price over-cost by Vega(K, σ_ATM)—producing clean quadratic log-moneyness wings, though missing exact pillar recovery by a few bps. Exact Price Vanna-Volga adds the exact market-minus-BS cost of the 3 pillars and inverts Black-Scholes, hitting all 3 pillars to 0.00 bps!

Publication record

Published
November 13, 2026
Updated
November 14, 2026
Author
Dr. Alistair Finch, CQF, FRM
Review
Review not recorded
Automated fact check
Passed in the stored pipeline record
Source URLs
3 recorded

AI assistance recorded: anthropic/claude-opus-4.6. This field does not establish human review or source verification.

Academic & Institutional Attribution

Cite This Research & Embed Live Data Card

Open Attribution License (CC BY 4.0)

Standard academic & institutional research paper citation format.

Dr. Alistair Finch, CQF, FRM. (2026, November 14). Malz Quadratic vs Vanna-Volga FX Volatility Smile Interpolation. VolSurface Desk. https://volsurfacedesk.com/desk-notes/malz-quadratic-vs-vanna-volga-fx-smile-interpolation
Explore FX Risk Reversal Strangle Surfaces Hub (20) →

VolSurface Desk Data, Benchmarks & Contributor Records

Malz Quadratic vs Vanna-Volga FX Smile Interpolation Guide | VolSurface Desk