💰 quantitative finance

Option pricing model under the G-expectation framework

This paper develops a unified risk-neutral valuation framework under the G-expectation environment that yields a nonlinear G-Black-Scholes equation, and proposes efficient explicit and implicit finite difference schemes based on a logarithmic transformation to accurately and stably compute option prices under model uncertainty.

Original authors: Ziting Pei, Xingye Yue, Xiaotao Zheng

Published 2026-03-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ziting Pei, Xingye Yue, Xiaotao Zheng

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are a financial architect trying to design a safety net (an "option") for a client. In the perfect world of traditional finance, the weather (stock prices) follows a predictable pattern, like a gentle, steady breeze. You can use a standard blueprint (the famous Black-Scholes formula) to calculate exactly how much that safety net should cost.

But in the real world, the weather is chaotic. Sometimes the wind is calm; other times, it's a hurricane. We don't know exactly how wild the storm will get, only that it will stay within a certain range. This is uncertainty.

This paper is about building a better, more robust blueprint for that safety net when the weather is unpredictable. Here is the story of how they did it, broken down into simple concepts:

1. The Problem: The "Perfect Storm" Assumption

Traditional models assume the stock price moves like a calm boat on a lake. But in reality, the "wind" (volatility) changes. Sometimes it's light, sometimes it's heavy.

The Old Way: Assume the wind is always the same speed. If you're wrong, you lose money.
The New Way (G-Expectation): Assume the wind speed is unknown but stays within a specific "safe zone" (e.g., between a gentle breeze and a gale). The goal is to price the option so you are safe even in the worst-case storm within that zone.

2. The Solution: A New Mathematical Compass

The authors created a new equation (the G-Black-Scholes equation) that accounts for this uncertainty. Instead of giving you one single price, it calculates a price that protects you no matter how the volatility behaves within that range.

However, solving this new equation is like trying to navigate a ship through a foggy, stormy sea using a very complex, non-linear map. It's mathematically heavy and computationally expensive (it takes a long time for computers to solve).

3. The Magic Trick: The "Logarithmic Transformation"

Here is the paper's biggest innovation. Imagine you are trying to measure a mountain.

The Hard Way: You measure the mountain in raw feet from sea level. The numbers are huge, and the slope changes drastically at the top, making it hard to calculate step-by-step.
The Smart Way (Logarithmic Transformation): Instead of measuring raw height, you measure the ratio of height. You zoom out. Suddenly, the steep, jagged mountain looks like a gentle, smooth hill.

The authors took the stock price and applied a "logarithmic lens" to it.

Why it matters: In the original "raw" view, the computer has to take tiny, cautious steps to avoid crashing (stability issues). In the "log" view, the computer can take bigger, faster steps without losing its balance.
The Result: They found that this transformation makes the computer 63% faster and much more efficient, while still giving the exact same accurate answer.

4. The Two Engines: Explicit vs. Implicit

To solve these equations, the authors built two different "engines" (algorithms):

The Explicit Engine (The Sprinter): This method looks at the current state and predicts the next step instantly. It's very fast per step, but it's fragile. If the steps are too big, it trips and falls. The authors proved that with their "logarithmic lens," this sprinter can run much faster without tripping.
The Implicit Engine (The Climber): This method looks at the current step and the future step simultaneously to ensure stability. It's slower per step because it has to do extra math (solving a puzzle at every step), but it never falls, no matter how big the steps are. The authors proved that their version of this climber is also reliable and converges to the right answer.

5. The Proof: Testing on Real Scenarios

They didn't just do the math on paper; they tested it with two classic financial scenarios:

The Butterfly Spread: A complex bet where the payoff looks like a butterfly's wings (it goes up, then down, then up). This is a "bumpy" road.
The Digital Call: A bet that pays a fixed amount if the stock crosses a line, and nothing otherwise. This is a "cliff" (a sudden jump).

The Results:

Both engines worked perfectly, finding the correct price.
The "Logarithmic Transformation" made the "Sprinter" (Explicit method) significantly more efficient.
Even when the road was bumpy or had a cliff (discontinuous payoffs), the methods held up, proving they are robust tools for real-world uncertainty.

The Bottom Line

This paper gives financial engineers a better map and a faster car for navigating uncertain markets.

The Map: A new framework (G-expectation) that handles unknown volatility realistically.
The Car: A mathematical trick (logarithmic transformation) that lets computers solve these complex problems much faster and cheaper than before.

In short: They figured out how to price financial safety nets in a chaotic world without spending a fortune on computer time to do it.

1. Problem Statement

The paper addresses the challenge of pricing financial derivatives in markets characterized by model uncertainty, specifically where volatility is not constant but lies within a known range. Traditional Black-Scholes models assume constant volatility and a linear probability framework, which often fails to capture real-world market risks.

The authors focus on the G-expectation framework, a sublinear expectation theory developed by Peng that models uncertainty without relying on a single probability measure. In this setting:

The underlying asset follows a G-geometric Brownian motion.
The pricing problem leads to a fully nonlinear Partial Differential Equation (PDE) (the G-Black-Scholes equation) rather than the linear PDE of the classical model.
The goal is to develop robust, accurate, and computationally efficient numerical methods to solve this nonlinear PDE for option valuation.

2. Methodology

A. Theoretical Framework

Risk-Neutral Pricing under G-Expectation: The authors define risk neutrality in the G-framework, where the G-expected return of all risky securities equals the risk-free rate $r$ .
Derivation of the G-Black-Scholes Equation: Using the nonlinear Feynman-Kac formula, they derive a fully nonlinear PDE for the option price $U(t, S)$ :
$\frac{\partial U}{\partial t} + rS \frac{\partial U}{\partial S} + \frac{1}{2} \sup_{\Sigma \in [\underline{\Sigma}, \overline{\Sigma}]} \left( \sigma^2 S^2 \Sigma^2 \frac{\partial^2 U}{\partial S^2} \right) - rU = 0$
Here, $\Sigma$ represents the uncertain volatility within the interval $[\underline{\Sigma}, \overline{\Sigma}]$ . The "sup" operator makes the equation fully nonlinear.

B. Logarithmic Transformation

To improve numerical performance, the authors introduce a change of variables: $X = \ln(S)$ .

This transforms the original PDE into an alternative nonlinear PDE for $V(t, X)$ :
$\frac{\partial V}{\partial t} + r \frac{\partial V}{\partial X} + \sup_{\Sigma \in [\underline{\Sigma}, \overline{\Sigma}]} \left[ \frac{1}{2}\sigma^2 \Sigma^2 \left( \frac{\partial^2 V}{\partial X^2} - \frac{\partial V}{\partial X} \right) \right] - rV = 0$
Significance: This transformation converts the variable coefficients (dependent on $S$ ) into constant coefficients, significantly relaxing the stability constraints for explicit numerical schemes.

C. Numerical Schemes

The authors propose two finite-difference schemes to solve the transformed PDE:

Explicit Finite Difference Scheme: A direct discretization where the solution at the next time step is calculated directly from the current step.
Implicit Finite Difference Scheme: A scheme where the solution at the next time step depends on itself, requiring the solution of a nonlinear algebraic system. To handle the nonlinearity, they employ a Picard iteration (inner loop) at each time step.

3. Key Contributions

Unified Risk-Neutral Framework: Formulated a comprehensive risk-neutral valuation framework within the G-expectation setting, generalizing the Black-Scholes model to handle volatility uncertainty.
Rigorous Theoretical Analysis:
- Proved that both explicit and implicit schemes are consistent, stable, and monotone.
- Established convergence to the viscosity solution of the nonlinear PDE based on the Barles-Souganidis theorem.
- Provided convergence rate estimates ( $O(\Delta t^{\beta/2} + h^\beta)$ ) for both schemes.
- Proved the monotonic convergence of the inner iteration for the implicit scheme.
Optimization via Logarithmic Transformation: Demonstrated theoretically and numerically that the logarithmic transformation relaxes the Courant-Friedrichs-Lewy (CFL) stability condition for explicit schemes, allowing for larger time steps and reduced computational cost.
Algorithmic Efficiency: Showed that the implicit scheme's inner iteration converges rapidly (typically within 2 iterations per time step), making it a viable alternative to explicit methods despite the nonlinearity.

4. Results

The authors validated their methods using numerical experiments on two types of options:

Butterfly Spread: A non-convex payoff function.
Digital Call Option: A discontinuous payoff function (challenging for numerical stability).

Key Findings:

Accuracy: Both schemes achieved approximately second-order accuracy for smooth payoffs (Butterfly spread) and near first-order accuracy for discontinuous payoffs (Digital call), consistent with theoretical expectations.
Stability Constraints:
- The explicit scheme on the original $S$ -domain requires a time step $\Delta t \propto 1/S_{max}^2$ .
- The explicit scheme on the transformed $X$ -domain requires $\Delta t \propto 1/(\ln S_{max} - \ln S_{min})^2$ .
- Result: The logarithmic transformation reduced the required number of time steps by approximately 63% (e.g., from 22,500 to 8,286 steps for a specific grid size), drastically lowering CPU time.
Implicit Scheme Performance: The implicit scheme was unconditionally stable (no strict mesh ratio condition) and the inner iteration converged in very few steps (average of 2), confirming its efficiency.
Comparison: The log-transformed explicit method offered the best balance of speed and accuracy, while the implicit method provided robustness for very fine grids without stability constraints.

5. Significance

This paper makes a significant contribution to quantitative finance and numerical analysis in the following ways:

Robustness under Uncertainty: It provides a practical, mathematically rigorous tool for pricing options in incomplete markets where volatility is uncertain, moving beyond the limitations of the classical Black-Scholes model.
Computational Efficiency: The demonstration that a simple logarithmic transformation can drastically reduce computational costs for nonlinear PDEs is a valuable insight for practitioners implementing real-time pricing engines.
Theoretical Rigor: By establishing the convergence to viscosity solutions and proving stability/monotonicity, the paper ensures that the numerical results are not just approximations but mathematically sound solutions to the underlying financial problem.
Versatility: The proposed methods are applicable to a wide range of exotic and standard options, handling both convex and non-convex payoffs effectively.

In conclusion, the authors successfully bridge the gap between the abstract theory of G-expectation and practical numerical implementation, offering a superior framework for robust option pricing under model uncertainty.