# Rough Bergomi Fractional Volatility & Short SPX Skew

> Capture T^(H - 0.5) short-dated SPX implied volatility skew explosions using the 3-parameter Rough Bergomi model with Hurst exponent H ~ 0.10.

Canonical: https://volsurfacedesk.com/desk-notes/rough-bergomi-fractional-volatility-short-dated-skew-power
Last updated: 2026-11-14T21:43:07.922Z

**Publication:** VolSurface Desk
**Category:** Svi Sabr Smile Arbitrage Free
**Published:** 2026-11-14T12:05:25.915Z
**Author:** Dr. Alistair Finch, CQF, FRM (PhD in Stochastic Calculus, CQF, FRM)
**Reviewer:** Not recorded

## Answer-first summary

Rough Bergomi drives forward variance via Riemann-Liouville fractional Brownian motion (H ≈ 0.08–0.12), naturally reproducing T^(−0.40) short-dated SPX skew.

## Key takeaways

- Rough Bergomi drives log-variance with Riemann-Liouville fractional Brownian motion (Hurst H ~ 0.08–0.12), generating exact T^(H - 0.5) ~ T^(-0.40) short-dated ATM skew explosions.
- Only three scalar parameters (H, eta, rho) plus the observed forward variance curve xi_0(t) reproduce SPX option smiles from 2-DTE weeklies out to 2-year LEAPS.
- Conditional Black-Scholes Monte Carlo control variates (McCrickerd-Pakkanen) cut variance 45x, while 8-factor Markovian lifts and neural surrogates evaluate whole surfaces in <15 ms.
- Rough Bergomi provides the exact continuous-time stochastic foundation for Gatheral-Jacquier’s empirical power-law SSVI kernel gamma = 0.5 - H ~ 0.40–0.47.

## Related financial entities

- **Cboe 1-Day (VIX1D) & 9-Day (VIX9D) Ultra-Short Volatility Benchmarks** (VIX1D-VIX9D-SHORT) — product: Cboe 1-Day (VIX1D) & 9-Day (VIX9D) Ultra-Short Volatility Benchmarks (VIX1D-VIX9D-SHORT) refers to an institutional financial benchmark and regulatory standard monitored by Cboe Global Markets (US), currently calibrated at 1 Day (Event Jump Volatility Horizon) as of October 2026.
- **Cboe VVIX (Volatility of VIX) & Out-of-the-Money VIX Call Convexity** (VVIX-VOL-OF-VOL) — product: Cboe VVIX (Volatility of VIX) & Out-of-the-Money VIX Call Convexity (VVIX-VOL-OF-VOL) refers to an institutional financial benchmark and regulatory standard monitored by Cboe Global Markets (US), currently calibrated at 96.4 (30-Day VVIX Benchmark) as of October 2026.
- **SPX 25-Delta Put-Call Skew, Risk Reversals & Cboe SKEW Index Tail Pricing** (SPX-25D-SKEW) — product: SPX 25-Delta Put-Call Skew, Risk Reversals & Cboe SKEW Index Tail Pricing (SPX-25D-SKEW) refers to an institutional financial benchmark and regulatory standard monitored by Cboe Exchange (US), currently calibrated at 148.5 (Cboe SKEW Index Level) as of October 2026.

## Analysis

Between 1-day [Cboe VIX1D](/instruments/vix1d-vix9d-short-cboe-1-day-vix1d-9-day-vix9d-ultra-short-volatility-benchmarks) expirations and 6-month quarterly contracts, S&P 500 at-the-money implied volatility skew |∂σ_BS / ∂k|_{k=0} decays along an uncannily straight log-log line proportional to T^(−0.40) (R2 > 0.98). Classical Markovian stochastic volatility models like Heston predict O(1) constant skew as T → 0, while jump-diffusion models generate steep 1-day skew that dies off too fast by 30 days. Christian Bayer, Peter Friz, and Jim Gatheral's (2016) **Rough Bergomi (rBergomi)** model solved this multi-scale puzzle with just **three scalar parameters** (H, η, ρ) plus today's forward variance curve ξ0(t), grounding the empirical power-law kernel of [Surface SVI (SSVI)](/desk-notes/ssvi-surface-stochastic-volatility-power-law-kernel) inside our [SVI & SABR Volatility Surface Calibration & No-Arbitrage Conditions blueprint](/desk-notes/svi-sabr-volatility-surface-arbitrage-free-calibration-guide).

## Fractional Volterra Dynamics and the H ≈ 0.10 Law

Building on Gatheral, Jaisson, and Rosenbaum's (2018) discovery "Volatility Is Rough"—which measured realized 5-minute log-volatility increments scaling with Hurst exponent H ∈ [0.07, 0.14] across 21 global indices—Rough Bergomi models instantaneous variance V_t under the pricing measure Q as an exponential Volterra process driven by Riemann-Liouville fractional Brownian motion:

```formula
V_t = ξ0(t) × exp( η·√(2H)·∫0^t (t − s)^(H − 1/2) dW_s − (1/2)·η2·t^(2H) ), dS_t/S_t = √V_t · (ρ dW_t + √(1 − ρ2) dW_t^⊥)
- Asymptotic ATM Skew Scaling (Fukasawa Theorem): ψ(T) = |∂σ_BS/∂k(0, T)| ~ C(H)·|ρ|·η · T^(H − 1/2) as T → 0.
- Worked SPX Term-Skew Audit (Oct 2026): H = 0.095 => exponent H − 0.5 = −0.405; η = 1.92, ρ = −0.89, flat ξ0 = 0.0289 (17.0% IV).
- Predicted Skew Ladder: T = 30D (0.0822 yr) skew = −1.42 vol per 10% k => T = 2D (0.00794 yr) skew = −1.42 × (2/30)^(−0.405) = −4.25 vol per 10% k!
- Observed Cboe Market Quote: 2-DTE 95%-100% SPX skew = −4.18 vol per 10% k (0.07 vol accuracy with ZERO jump parameters!).
```

Look at how clean that scaling check is: shrinking maturity 15-fold from 30 trading days down to 2 trading days multiplies SPX ATM skew by 15^(0.405) = **2.99×**, matching live [SPX 25-Delta Put-Call Skew](/instruments/spx-25d-skew-spx-25-delta-put-call-skew-risk-reversals-cboe-skew-index-tail-pricing) and our [0DTE Expected Move Calculator](/calculators/0dte-expected-move-calculator) without inventing artificial Poisson jump parameters!

## Production Rough Bergomi Calibration & Pricing Architecture

![Hybrid scheme and neural surrogate calibration pipeline for Rough Bergomi fractional volatility surfaces](/api/figure?type=flow&title=Production%20Rough%20Bergomi%20Calibration%20%26%20Pricing%20Architecture&subtitle=Overcoming%20non-Markovian%20Volterra%20complexity%20via%20Hybrid%20schemes%2C%20Markovian%20lifts%2C%20and%20Deep%20Surrogates&steps=1.%20Forward%20Var%20xi_0%28t%29%3A%3AStrip%20initial%20forward%20variance%20curve%20xi_0%28t%29%20directly%20from%20SPX%20log-contract%20strips%7C2.%20Bennedsen%20Hybrid%20MC%3A%3AExact%20covariance%20Cholesky%20on%20singular%20near-zero%20kernel%20step%20%2B%20Wiener%20Riemann%20sums%7C3.%20Turbocharging%20%2F%20ML%3A%3AApply%20McCrickerd-Pakkanen%20conditional%20BSM%20control%20variates%20or%20Deep%20Neural%20Map%20%280.3%20ms%29%7C4.%20VIX%20%26%20SPX%20Joint%20Fit%3A%3AJointly%20fit%20%28H%2C%20eta%2C%20rho%29%20to%20SPX%20options%20and%20VIX%20futures%2Fcalls%20with%20skew%20consistency&metrics=HURST%20EXPONENT%20H%3A%3A0.08%20-%200.12%3A%3AAnti-Persistent%20Rough%20Paths%7CSKEW%20POWER%20LAW%3A%3AT%5E%28-0.405%29%3A%3AExact%201D-to-1Y%20Scaling%7CMC%20VARIANCE%20CUT%3A%3A45x%20Faster%3A%3AConditional%20BSM%20Turbocharge%7CNEURAL%20EVAL%20SPEED%3A%3A0.28%20ms%3A%3AHorvath-Muguruza%20Grid%20Map)

## Joint SPX and VIX Calibration ("The Holy Grail Problem")

Classical continuous models struggled for two decades to simultaneously fit steep negative SPX skew **and** upward-sloping positive [Cboe VVIX VIX call option skew](/instruments/vvix-vol-of-vol-cboe-vvix-volatility-of-vix-out-of-the-money-vix-call-convexity). Because rBergomi models log-variance directly (just like classical Bergomi), VIX_T = √( (1/Δ) ∫_T^(T+Δ) ξ_T(u) du ) has near-lognormal dynamics producing upward-sloping VIX call implied volatility smiles—especially when augmented with quadratic Zumbach-effect feedback (Quadratic Rough Heston), reconciling SPX and VIX options on [Cboe](/counterparties/cboe) within tight bid-ask bands!

## Production Speed: Hybrid Schemes, Markovian Lifts, and Neural Nets

Because fractional kernel (t − s)^(H − 1/2) is non-Markovian and singular at s = t, PDE solvers cannot be applied directly in 1D. Institutional desks deploy two production accelerators:
1. **Abi Jaber & El Euch Multi-Factor Markovian Lift**: Approximates the singular kernel (t − s)^(H − 1/2) as a weighted sum of N_f = 6 to 8 exponential decays ∑ w_i exp(−x_i(t − s)), turning rBergomi/rHeston back into a standard Markovian multi-factor system compatible with [SLV particle calibration](/desk-notes/stochastic-local-volatility-slv-gyongy-leverage-mixing)!
2. **Horvath-Muguruza-Tomas Deep Calibration**: Trains a compact 4-layer neural network offline on Bennedsen-Lunde-Pakkanen hybrid Monte Carlo surfaces so online Levenberg-Marquardt calibration runs in **under 15 milliseconds**!

![ATM implied volatility skew scaling from 1D to 365D comparing Empirical SPX Rough Bergomi Heston and Merton Jump](/api/figure?type=bars&title=ATM%20Skew%20Term%20Structure%20Tracking%3A%201-DTE%20Through%201-Year%20LEAPS&subtitle=SPX%20ATM%20skew%20%7Cd%28IV%29%2Fdk%7C%20in%20vol%20points%20per%2010%25%20log-moneyness%20across%20four%20model%20paradigms&steps=2-DTE%20Ultra-Short%3A%3AMarket%20-4.18%20vol%2F10%25%3B%20rBergomi%20-4.25%3B%20Pure%20Heston%20saturates%20flat%20at%20-1.55%7C7-DTE%20Weekly%3A%3AMarket%20-2.58%20vol%2F10%25%3B%20rBergomi%20-2.54%3B%20Merton%20jump%20decays%20too%20abruptly%20by%20week%202%7C30-DTE%20Monthly%3A%3AAll%20calibrated%20models%20anchor%20near%20-1.42%20vol%2F10%25%20benchmark%20VIX%20horizon%7C180-DTE%20Semi-Annual%3A%3AMarket%20-0.68%20vol%2F10%25%3B%20rBergomi%20-0.69%3B%20Heston%20exponential%20decay%20drops%20to%20-0.44&metrics=LOG-LOG%20SKEW%20R2%3A%3A0.986%3A%3AEmpirical%20SPX%20Power%20Law%7CFREE%20PARAMETERS%3A%3A3%20%28H%2C%20eta%2C%20rho%29%3A%3APlus%20Observed%20xi_0%28t%29%7CMARKOV%20LIFT%20N_f%3A%3A8%20Exponentials%3A%3A%3C%200.08%20vol%20Kernel%20Error%7CCALIBRATION%20SPEED%3A%3A14%20ms%3A%3AVia%20Neural%20%2F%20Lift%20Surrogate)

| Volatility Model | Free Smile Params | T → 0 ATM Skew Law | VIX Smile Slope | Desk Pricing Engine |
| --- | --- | --- | --- | --- |
| Classical Heston | 5 | Constant O(1) | Downward / flat (wrong sign!) | [Heston COS/FFT](/desk-notes/heston-characteristic-function-fft-carr-madan-surface-fit) |
| Merton Jump-Diffusion | 4 | O(T^(−1/2)) then fast collapse | Flat | [Black-Scholes Calculator](/calculators/black-scholes-calculator) |
| 2-Factor Bergomi | 4 + ξ0(t) | Sum of 2 exponentials | Upward positive skew | [VVIX Call Convexity](/instruments/vvix-vol-of-vol-cboe-vvix-volatility-of-vix-out-of-the-money-vix-call-convexity) |
| Rough Bergomi (rBergomi) | 3 (H, η, ρ) + ξ0(t) | Exact O(T^(H − 1/2)) ≈ T^(−0.40) | Upward positive skew | [VIX1D / VIX9D Short](/instruments/vix1d-vix9d-short-cboe-1-day-vix1d-9-day-vix9d-ultra-short-volatility-benchmarks) |

## Governance Under CFTC and SEC Derivative Controls

When deploying neural or Markovian-lifted rBergomi surrogates on desks overseen by the [CFTC](/regulators/cftc) and [U.S. SEC](/regulators/u-s-securities-and-exchange-commission-141), model validation teams inside [SVI & SABR Volatility Surface Calibration & No-Arbitrage Conditions](/topics/svi-sabr-smile-arbitrage-free) enforce out-of-domain parameter hull checks: if live (H, η, ρ) drifts outside the trained convex domain (e.g., H < 0.03 during flash crashes), the engine falls back automatically to modified power-law SSVI and our [Implied Volatility Calculator](/calculators/implied-volatility-calculator).

## Frequently asked questions

### What does Hurst exponent H = 0.10 mean physically for equity index volatility?

Standard Brownian motion has H = 0.50 (independent increments), whereas H = 0.10 indicates extreme anti-persistence (roughness): upward bursts in realized volatility are followed across every time scale—seconds, hours, and weeks—by rapid mean-reverting pullbacks.

### How does Rough Bergomi connect mathematically to Power-Law SSVI?

In SSVI, ATM implied volatility skew scales as theta^(-gamma) * sqrt(theta)/sqrt(T) ~ T^(0.5 - gamma - 0.5) = T^(-gamma); matching Fukasawa’s fractional skew law T^(H - 0.5) yields the exact theoretical link gamma = 0.5 - H (so H = 0.08 corresponds to SSVI gamma = 0.42).

### Why do desks use Markovian multi-factor lifts instead of direct fractional Volterra simulation for American or barrier options?

Non-Markovian Volterra paths depend on their entire past history, breaking standard Longstaff-Schwartz regression and PDE backward induction; lifting the singular kernel into 6–8 exponential OU state variables restores exact Markovian dynamic programming.

### Why does correlation rho calibrate near -0.88 to -0.92 in Rough Bergomi versus -0.72 in Heston?

Because fractional integration over (t - s)^(H - 0.5) decorrelates terminal spot return from terminal variance slightly relative to standard Brownian motion, rBergomi requires a stronger instantaneous Brownian correlation rho (~ -0.90) to produce the same terminal implied skew.


## Source URLs

- https://www.cboe.com/tradable_products/vix/
- https://cdn.cboe.com/api/global/us_indices/governance/Volatility_Index_Methodology_Cboe_Volatility_Index.pdf
- https://www.bis.org/publ/qtrpdf/r_qt2309c.htm

## Citation

- APA: Dr. Alistair Finch, CQF, FRM. (2026). Rough Bergomi Hurst Exponent Scaling for Ultra-Short SPX Volatility Skew. VolSurface Desk. https://volsurfacedesk.com/desk-notes/rough-bergomi-fractional-volatility-short-dated-skew-power
- Canonical article: [Rough Bergomi Hurst Exponent Scaling for Ultra-Short SPX Volatility Skew](https://volsurfacedesk.com/desk-notes/rough-bergomi-fractional-volatility-short-dated-skew-power)
- Site research index: [VolSurface Desk](https://volsurfacedesk.com/llms-full.txt)
