Global outstanding equity-linked OTC derivatives totaled $8.7 trillion notional at mid-2024 (BIS data), with institutional volatility arbitrage representing a significant subset of this market. The majority of vol arb desks run delta-hedged option portfolios calibrated against stochastic volatility models. The Heston (1993) framework remains the industry standard because it explicitly captures what Black-Scholes ignores: volatility clustering, mean reversion, and the leverage effect — all of which directly drive P&L in options trading books.
The Market Structure: Why Constant Volatility Models Fail
Black-Scholes assumes volatility is constant. Market reality contradicts this assumption systematically. During the March 2020 drawdown, VIX closed at 82.69 on March 16 — the highest closing price ever recorded, surpassing the previous record of 80.86 from November 2008 — while realized volatility exhibited violent mean reversion within weeks. Funds running Black-Scholes-calibrated hedges experienced significant losses as their models failed to anticipate volatility dynamics across strikes and maturities.
The volatility smile — the characteristic skew in implied volatility across option strikes — cannot be explained by constant volatility models. Heston solves this by modeling volatility itself as a mean-reverting stochastic process with four critical parameters:
κ (kappa): Mean reversion speed — how quickly volatility returns to long-term average
θ (theta): Long-run variance level — the equilibrium volatility target
σ_v (sigma): Volatility-of-volatility — determines smile curvature and tail risk
ρ (rho): Correlation between spot returns and volatility — captures the leverage effect
For equity markets, calibrated parameters typically show ρ between -0.6 and -0.8 across major indices, meaning volatility rises when stocks fall — though specific values vary by asset, period, and calibration methodology. This negative correlation is fundamental to skew pricing.
Trade Structure: Volatility Arbitrage via Model Calibration
Position Construction:
Volatility arbitrage desks typically structure trades as long OTM options hedged with short ATM positions, maintained delta-neutral through continuous rebalancing.
Execution Process:
Calibrate Heston to current surface: Fit model parameters (κ, θ, σ_v, ρ, v₀) to market option prices using least-squares optimization. Calibration typically minimizes the weighted sum of squared differences between model and market prices across multiple strikes and maturities.
Identify mispricings: Compare Heston fair values to market quotes. Deviations exceeding 2–3% signal trading opportunities, particularly in the wings where σ_v impacts pricing most.
Execute skew trades: When market overprices downside protection (implied ρ too negative), desks sell put spreads and buy volatility in out-of-the-money regions where convexity is mispriced.
Delta hedge continuously: Maintain spot-neutral exposure while harvesting gamma as the underlying moves. Hedging frequency correlates with realized volatility — higher vol requires more frequent rebalancing.
Position Sizing: Vega-weighted to maintain consistent volatility exposure (typically $50–100k per vega point), diversified across 20–30 single stocks to reduce idiosyncratic risk.
P&L Mechanics: Three Sources of Returns
1. Gamma Scalping
Delta-hedge at frequency tied to realized volatility. Profit from each rebalancing cycle equals:
P&L = 0.5 × Γ × (ΔS)²
When realized volatility exceeds implied volatility at entry, gamma scalping generates positive returns that exceed theta decay. As an illustrative example, on an equity options portfolio with $10M gamma exposure, a 1% realized volatility edge can translate to approximately 15–20 basis points daily, though actual results depend on rebalancing frequency, transaction costs, and bid-ask spreads.
2. Vega Convergence
When market reprices options toward Heston fair value, vega P&L materializes:
P&L = Vega × Δσ_implied
Mispricings in liquid markets typically correct within 5–10 trading days under normal conditions, yielding 30–50 bps per position — though convergence speed varies with market liquidity and whether other participants recognize the same mispricing. In stressed or illiquid markets, convergence can take significantly longer.
3. Correlation Risk Capture
Heston’s ρ parameter predicts how volatility responds to spot moves. Properly modeling this correlation prevents higher-order Greeks (vanna, volga) from exploding during market stress. During Q1 2020, funds that ignored correlation dynamics in their hedging experienced catastrophic losses as volatility skew steepened dramatically.
Critical Risk: Model specification error. If Heston calibration targets only ATM options, wing prices will be systematically wrong. The volatility-of-volatility parameter (σ_v) is particularly challenging to estimate — underestimate it, and you underprice tail risk; overestimate, and you overpay for protection that won’t materialize.
The Quant Edge: Volatility-of-Volatility Drives Smile Curvature
σ_v determines everything about tail pricing. Higher vol-of-vol means fatter tails and more expensive OTM options. Typical calibrated ranges observed by practitioners:
Stable markets: σ_v ≈ 0.3–0.5
Crisis periods: σ_v > 1.5
Time-varying σ_v models demonstrably reduce hedging error versus constant-parameter fits. Funds that dynamically recalibrate daily capture regime shifts faster, extracting alpha from both directional (gamma) and structural (vega) sources.
Mean reversion speed (κ) dictates term structure arbitrage opportunities. Higher κ implies faster volatility convergence, meaning shorter-dated options become relatively expensive versus longer maturities. This creates calendar spread opportunities that Black-Scholes models miss entirely because they cannot capture the term structure of volatility.
Practical Implementation Insight: Calibration challenges arise because multiple parameter sets can fit market prices equally well. Professional desks address this by:
Constraining parameters to economically reasonable ranges based on historical estimates
Using variance swap prices to anchor θ (long-term variance)
Regularizing the calibration objective to prevent overfitting
Implementing multi-stage optimization: global search followed by local refinement
Key Takeaway: Model Correctness Is the Trade
The edge in volatility arbitrage isn’t predicting whether volatility will rise or fall. It’s about pricing second-order Greeks correctly when everyone else is using simplified models.
Heston enables funds to express views on:
Skew mispricing (exploiting incorrect ρ assumptions)
Term structure arbitrage (trading κ and θ dynamics)
Convexity capture (monetizing σ_v calibration errors)
Volatility arbitrage isn’t about being long or short volatility — it’s about being long model superiority. Funds that systematically price options more accurately than the market earn consistent risk-adjusted returns regardless of volatility direction.
References & Further Reading
Heston, S. L. (1993). “A Closed-Form Solution for Options with Stochastic Volatility with Applications to Bond and Currency Options.” Review of Financial Studies, 6(2), 327–343. DOI: 10.1093/rfs/6.2.327
Bank for International Settlements (2024). “OTC Derivatives Statistics at End-June 2024.” Statistical Release, November 2024.
SIFMA Research (2020). “The VIX’s Wild Ride: Covid-19 Crisis Analysis.” Market Analysis Report, April 2020.
International Swaps and Derivatives Association (2024). “Key Trends in the Size and Composition of OTC Derivatives Markets in the First Half of 2024.” December 2024.
Gatheral, J. (2006). The Volatility Surface: A Practitioner’s Guide. John Wiley & Sons. [Industry standard reference on volatility modeling]
Bergomi, L. (2015). Stochastic Volatility Modeling. Chapman and Hall/CRC Press. [Advanced treatment of stochastic vol models]
Christoffersen, P., Heston, S., & Jacobs, K. (2009). “The Shape and Term Structure of the Index Option Smirk: Why Multifactor Stochastic Volatility Models Work So Well.” Management Science, 55(12), 1914–1932.
Data Sources: Market statistics verified through Bank for International Settlements, ISDA, and regulatory filings (OCC, CFTC) as of H1 2024. Parameter ranges (ρ, σ_v) and P&L examples reflect typical calibrated values and practitioner observations; specific values vary by asset, market regime, and methodology. Heston model implementation details reflect standard institutional practices documented in quantitative finance literature 1993–2024.
About This Analysis: Written for quantitative researchers, hedge fund analysts, and institutional traders. All claims are evidence-backed with academic and regulatory sources. No theoretical speculation — only verified market mechanics and implementation practices used by professional volatility arbitrage desks.
Cover photograph: Warren LeMay, CC BY-SA 2.0, via Wikimedia Commons.