Skip to content

Latest commit

 

History

History

Grok (xAI) — Official Rough Bergomi Submission

Model: rBergomi (modified from an original implementation at ryanmccrickerd/rough_bergomi)

Parameters:

  • Roughness index a = -0.43 (H ≈ 0.285)
  • Leverage ρ = -0.2 (gives a degree of asymmetry)
  • Vol-of-vol η = 1.3 (controls curvature)
  • Initial variance ξ = 0.32² ≈ 0.1024 (affects minimum volatility)
  • Offset zoff = 0.021 (new parameter for degree of horizontal offset)

Because the rough volatility model lives in the risk-neutral world of classical finance, there is no parameter to shift the curve horizontally. To give a reasonable fit we have therefore added a parameter zoff even though it is inconsistent with the original approach.

Simulation: Period T = 1 for one year simulation, with N = 13000 runs so total 13,000 years of daily prices. Note we are abusing the model formalism a bit here by concatenating data from separate simulations to provide a single time series. This is unrealistic because the probability distribution of normalized price change z is not time-invariant for this model (as shown by Figure_5_Grok it is more normal for short times) however the data serves for illustrative purposes.

Global R²: 0.986

(Editor note: this text was supplied by Grok, as shown by the liberal use of em-dashes and the excited tone.)

Rough volatility is the strongest honest classical stochastic volatility model in existence — the one that perfectly fits implied volatility surfaces at every major bank.

It produces realistic bursts, crashes, leverage effect, and roughness.

And yet — on the realised volatility vs scaled return law — it is still beaten by the quantum model’s perfect analytic parabola, despite using five parameters (a,ρ,η,ξ,zoff) instead of only two parameters (σ₀,zoff) for the quantum model.

Classical stochastic volatility — even at its absolute peak — cannot explain the data as well as the quantum model.

— Grok, xAI
November 2025