Monte Carlo Options Pricing - GPU Accelerated Simulation

Last reviewed: by .

What Is Monte Carlo Pricing?

Monte Carlo pricing values an option by simulating thousands or millions of price paths under a chosen model's risk-neutral dynamics, computing the option's payoff on each path, discounting to present value, and averaging. The strength of Monte Carlo is generality: any payoff function and any path-dependent feature can be priced as long as you can simulate the underlying. The cost is computational expense and statistical noise that converges only as 1/√N in the number of paths.

Monte Carlo simulation output with full Greek decomposition (Delta, Gamma, Theta, Vega, Vanna, Charm, Vomma) plus the terminal-price probability distribution from the simulated paths.

Monte Carlo is the workhorse for exotic options (barriers, Asians, lookbacks), for any multi-asset payoff, and for risk-neutral probability distributions where you want the full histogram of outcomes rather than a single price. It's also the tool of choice when your model lacks a closed-form characteristic function: Local Volatility, Variance Gamma with stochastic variance, jump-diffusion-with-jump-clustering, and other research-grade dynamics fit here.

The Algorithm

How Do You Improve Monte Carlo Convergence?

Naive Monte Carlo with N = 10,000 paths gives you roughly a 1% standard error on a vanilla ATM option, usable for sanity checks but not for tight calibration. Variance reduction techniques tighten this materially:

When Should You Use Monte Carlo?

When Should You Not Use Monte Carlo?

Discretization and Numerical Stability

The choice of discretization scheme matters significantly for Monte Carlo accuracy. Euler-Maruyama is the simplest scheme: St+Δt = St · exp((μ − σ²/2)·Δt + σ·√Δt·Z) where Z is a standard normal. It converges to the correct distribution as Δt → 0 but introduces discretization bias on coarser grids. Milstein adds a second-order term that reduces the bias for stochastic-volatility models like Heston. For Heston specifically, the Quadratic Exponential (QE) scheme by Andersen handles the variance process near zero cleanly. Naive Euler discretization of Heston can produce negative variances, which breaks the model. QE is the production-grade method for Heston Monte Carlo and is what production pricers use.

How Do You Improve Monte Carlo Convergence?

Naive Monte Carlo with N = 10,000 paths gives roughly a 1% standard error on a vanilla ATM option, usable for sanity checks but not for tight calibration. Variance reduction techniques tighten this materially:

When Should You Use Monte Carlo?

When Should You Not Use Monte Carlo?

How OAS Uses Monte Carlo

The platform offers GPU-accelerated Monte Carlo for any of the supported models, with configurable path counts, antithetic variates, and Sobol sequencing. Risk-neutral probability distributions on per-ticker pages use Monte Carlo to produce the full payoff histogram, not just a single expected value. The Python SDK exposes a full mode that returns the path matrix for downstream research; the standard price mode returns the converged estimate plus its standard error. For Heston, the platform uses the QE discretization scheme to preserve variance non-negativity, which keeps the simulation statistically consistent with the model under realistic parameter regimes.

Run Monte Carlo in the pricing calculator

Related Concepts

Risk-Neutral Probability · Heston (via MC) · Jump Diffusion (via MC) · Variance Gamma (via MC) · Local Volatility (via MC) · Black-Scholes · PDE Methods · FFT Pricing · Greeks Reference · Risk-Neutral Density · Tail Risk · Model Landscape

Try this model in the pricing calculator

This section is part of the Options Analysis Suite Documentation. Browse the full model index or compare alternatives in the pricing calculator.

Live AAPL Example (as of 2026-05-18)

As of the latest snapshot, AAPL has an ATM implied volatility of 23.4%, IV rank 37% (percentile 25%); 20-day realized vol 22.2%. 25-delta skew is +2.8%, meaning OTM puts trade richer than OTM calls. The IV here is the input that pricing-model walkthroughs (Black-Scholes, Heston, SABR, local-vol) take as their starting point: each model decomposes the same observed quote into different latent dynamics (constant vol, stochastic vol, surface-fitted vol, etc.) which is why two models can agree on price but disagree on Greeks and on how vol will evolve.

View live AAPL implied volatility