Deep dive into the 7 mathematical models behind options pricing: GBM, Heston, Jump Diffusion, GARCH, SABR, FBM, Bates — and the 3 pricers they pair with.
Why the Model Behind Your Simulation Matters
Every simulated stock chart is generated by a mathematical model — a set of equations that describe how prices evolve over time. The choice of model determines everything: whether the simulated market exhibits realistic volatility clustering, whether price jumps occur, whether the options chain produces a credible implied volatility smile, and ultimately whether your strategy test means anything at all.
Options Simulator uses seven stochastic price models organized into three tiers of increasing complexity, each paired with one of three professional option pricing engines . This is not an arbitrary architecture. Each pairing reflects a deliberate match between the assumptions of the price model and the capabilities of the pricer — the same logic used by quantitative trading desks when choosing their modeling stack.
This article explains each model's mathematical foundation, what real-world market behavior it captures, why it is paired with its specific option pricer, and when you should use it for strategy testing. We progress from the simplest (GBM) to the most complex (Bates), building intuition at each step about why additional complexity is sometimes necessary — and sometimes not.
Two Engines, One Simulation: How Price Models and Option Pricers Connect
Before diving into individual models, it is important to understand the dual-engine architecture that powers the simulator. Two independent but connected systems work together:
Engine 1 — The Price Generator creates the OHLC candlestick chart. When you click "Generate," the selected stochastic model produces an entire price history — hundreds of candles with realistic open, high, low, and close values. This happens once; the candles are then fixed for the duration of the simulation.
Engine 2 — The Option Pricer calculates the option chain. At every candle during playback, the pricer uses the current stock price, historical volatility, time to expiration, and a risk-free rate to calculate prices and Greeks (Delta, Gamma, Theta, Vega, Rho) for every strike and expiration. This happens in real time, every step.
The connection between them is configurational : each price model is assigned a specific option pricer based on mathematical consistency. Simpler models that assume constant volatility use faster, simpler pricers. Complex models that generate stochastic volatility or jumps require more sophisticated pricers that can handle these dynamics.
Tier Price Model Option Pricer Option Type
Essential GBM (Steady Drift) Black-Scholes European Jump Diffusion (Sudden Shock) Black-Scholes European Advanced Dynamics Heston (Shifting Volatility) Bjerksund-Stensland American FBM (Deep Tide) Bjerksund-Stensland American Professional Grade GARCH (Calm & Storm) Binomial Tree (CRR) American Bates (Perfect Storm) Binomial Tree (CRR) American SABR (Volatility Smile) Binomial Tree (CRR) American
The dual-engine architecture: each price generator is paired with a mathematically consistent option pricer The model-pricer pairing is configurable — but the defaults reflect mathematical consistency Tier 1: Essential — The Foundation
GBM — Geometric Brownian Motion (Steady Drift)
The seminal model. Geometric Brownian Motion is the mathematical engine behind the Black-Scholes-Merton framework that earned Myron Scholes and Robert Merton the 1997 Nobel Prize in Economics. Introduced in the context of option pricing by Black and Scholes in 1973, GBM assumes that stock prices follow a continuous random walk with a constant drift (expected return) and constant volatility.[1]
The model is defined by a stochastic differential equation (SDE): the instantaneous return has two components — a deterministic drift term proportional to the current price, and a random shock driven by a Wiener process (Brownian motion), also proportional to the current price. This multiplicative structure ensures prices cannot go negative and that returns are log-normally distributed.
Key parameters: drift (μ) — the expected annual return; volatility (σ) — the constant annual standard deviation of returns.
What it captures: Smooth, continuous price movement with a directional tendency. No sudden jumps. No volatility changes. Returns are normally distributed.
What it misses: Real markets exhibit fat tails (extreme events occur more often than the normal distribution predicts), volatility clustering (volatile periods follow volatile periods), and the leverage effect (volatility increases when prices fall). GBM captures none of these.
When to use it: GBM is your baseline. Test a strategy under "textbook" conditions first. If it doesn't work under GBM, adding complexity won't save it. GBM is also ideal for beginners learning the fundamental mechanics of options — how Delta changes with price, how Theta accelerates near expiration — without the confounding effects of volatile volatility.
Jump Diffusion — Merton Model (Sudden Shock)
Robert Merton extended GBM in 1976 by adding a jump component — sudden, discontinuous price movements that arrive randomly according to a Poisson process.[2]
Key parameters: drift (μ) and diffusion volatility (σ) as in GBM; plus jump intensity (λ) — average number of jumps per year; jump mean (μ_J) — average jump size; jump volatility (σ_J) — variability of jump sizes.
What it captures: Sudden price gaps caused by earnings announcements, regulatory decisions, or geopolitical shocks. The compound Poisson process generates the excess kurtosis (fat tails) observed in real return distributions that GBM cannot explain.
What it misses: Volatility is still constant between jumps. There is no volatility clustering or mean reversion. The jumps are independent of the current volatility state — in reality, jumps and high volatility tend to coincide.
When to use it: Test how your strategy handles sudden gaps. An iron condor that looks perfect under GBM may face catastrophic losses when a 5% overnight gap moves the underlying through a short strike. Jump Diffusion is the simplest way to introduce this risk.
Why Black-Scholes Pricing for Tier 1
Both GBM and Jump Diffusion assume constant volatility between shocks. The Black-Scholes pricing model makes the same assumption — it was derived specifically for GBM dynamics.[1]
Tier 2: Advanced Dynamics — Adding Realism
Heston — Stochastic Volatility (Shifting Volatility)
The Heston model, published by Steven Heston in 1993, represents a fundamental advance over GBM: volatility itself becomes a random variable governed by its own stochastic process.[3] correlated with the Brownian motion driving the stock price.
Key parameters: initial variance (v₀); long-run variance (θ) — the level to which variance reverts; mean-reversion speed (κ) — how fast variance returns to θ; volatility of volatility (ξ) — how much the variance itself fluctuates; correlation (ρ) — the relationship between price and volatility movements.
What it captures: The Heston model captures three empirically crucial phenomena. First, volatility clustering — periods of high volatility tend to persist, then gradually revert to normal levels. Second, the leverage effect — when ρ is negative (typically -0.7 to -0.9 in equity markets), falling prices coincide with rising volatility, and vice versa. Third, the volatility smile — the correlation between price and volatility movements produces the characteristic pattern where out-of-the-money puts have higher implied volatility than ATM options.[3]
Why it matters for options: The Heston model is widely accepted by practitioners precisely because it captures the volatility smile that is fundamental to real options markets.[4]
When to use it: Heston is the default choice for realistic options strategy testing. Any strategy that depends on implied volatility behavior — vertical spreads, iron condors, straddles, strangles — should be tested under Heston at minimum.
FBM — Fractional Brownian Motion (Deep Tide)
Standard Brownian motion has independent increments — what happens today is unrelated to what happened yesterday. Fractional Brownian Motion, formalized by Mandelbrot and Van Ness in 1968, relaxes this assumption by introducing long-range dependence controlled by a single parameter: the Hurst exponent (H).[5]
Key parameter: Hurst exponent (H), where 0 < H < 1. When H = 0.5, FBM reduces to standard Brownian motion (GBM). When H > 0.5, increments are positively correlated — upward movements tend to be followed by upward movements, creating persistent trends. When H < 0.5, increments are negatively correlated — movements tend to reverse, creating mean-reverting behavior.
What it captures: The empirical observation that financial markets exhibit long-term memory — trending behavior over medium horizons and mean reversion over longer horizons. This "fractional" memory structure cannot be captured by any Markov model (including GBM, Heston, and GARCH), which by definition have no memory.
What it misses: FBM does not model volatility dynamics separately. There are no jumps, no volatility clustering, and no smile effects.
When to use it: FBM is the right model for testing directional strategies — long calls, long puts, and momentum-based approaches — where the presence or absence of trend persistence is the key risk factor. Set H > 0.5 to test in a trending environment; set H < 0.5 to test in a mean-reverting one.
Why Bjerksund-Stensland Pricing for Tier 2
The Heston and FBM models generate price dynamics where volatility is no longer constant — it changes over time (stochastically in Heston, through long-memory in FBM). Black-Scholes, with its constant-volatility assumption, would be inconsistent. The Bjerksund-Stensland model provides an analytical approximation for American options that is fast enough for real-time option chain generation while handling non-constant volatility inputs.[6]
Tier 3: Professional Grade — Institutional Accuracy
GARCH — Generalized Autoregressive Conditional Heteroskedasticity (Calm & Storm)
The GARCH(1,1) model, developed by Tim Bollerslev in 1986 as a generalization of Robert Engle's ARCH model (1982, Nobel Prize 2003), captures volatility clustering through a fundamentally different mechanism than Heston.[7] [8]
Key parameters: ω — the base level of variance; α — the weight on recent squared returns (the "reactivity" to news); β — the persistence of previous variance; the constraint α + β < 1 ensures stationarity.
What it captures: The defining feature of GARCH is conditional heteroskedasticity — the variance of returns depends on past returns. After a large move (positive or negative), the model predicts elevated variance for subsequent periods. After calm periods, predicted variance decreases. This produces the characteristic pattern of volatility regimes: alternating periods of calm and storm.
What it misses: Standard GARCH(1,1) treats positive and negative returns symmetrically — a +5% return and a -5% return have the same impact on future variance. Real markets show asymmetry (the leverage effect), though extensions like EGARCH and GJR-GARCH address this.
When to use it: GARCH is ideal for testing strategies across changing volatility regimes. If you want to see how a covered call portfolio performs when a calm period suddenly transitions to high volatility — and then gradually calms down — GARCH produces this dynamic naturally.
Bates — Stochastic Volatility with Jumps (Perfect Storm)
The Bates model, introduced by David Bates in 1996, is the most comprehensive model in the simulator. It combines the stochastic volatility framework of Heston with the jump dynamics of Merton: the price follows a diffusion with time-varying, mean-reverting volatility (as in Heston), plus sudden jumps that arrive randomly according to a Poisson process (as in Jump Diffusion).[9]
Key parameters: All Heston parameters (v₀, θ, κ, ξ, ρ) plus all jump parameters (λ, μ_J, σ_J). This gives Bates the largest parameter space — and the greatest expressive power — of any model in the simulator.
What it captures: Everything the Heston model captures (volatility clustering, leverage effect, volatility smile) plus sudden price dislocations. Bates found that stochastic volatility alone could not fully explain the excess kurtosis observed in options-implied distributions — jumps were necessary to account for the market's pricing of extreme events.[9]
When to use it: Bates is your ultimate stress test. It generates the most challenging market conditions: a volatile environment punctuated by sudden shocks. If your strategy survives Bates simulation under the Market Crash scenario, it has passed a test that captures both the gradual buildup of fear (stochastic volatility rising) and the sudden realization of that fear (jump down).
SABR — Stochastic Alpha Beta Rho (Volatility Smile)
The SABR model, published by Hagan, Kumar, Lesniewski, and Woodward in 2002, was developed specifically to solve a problem that local volatility models could not: correctly predicting how the volatility smile moves when the underlying price changes .[10]
Hagan et al. demonstrated that local volatility models (à la Dupire) produce smile dynamics that are the opposite of observed market behavior. When prices fall, local vol models predict the smile shifts to higher prices — but in reality, the smile shifts in the same direction as the price. SABR, by modeling both the forward price and its volatility as correlated stochastic processes, captures the correct smile dynamics.
Key parameters: α — the initial volatility level; β — the elasticity parameter (0 ≤ β ≤ 1) that controls the backbone of the smile; ρ — the correlation between the forward price and its volatility; ν — the volatility of volatility.
What it captures: SABR is the industry standard for modeling the volatility smile in interest rate and FX options markets. Its closed-form implied volatility formula allows direct, analytical mapping between model parameters and the shape of the smile — including skew (controlled by ρ), curvature (controlled by ν), and the general slope of ATM volatility (controlled by β).
When to use it: SABR is the right model when the smile itself is the focus of your strategy. Butterfly spreads, ratio spreads, and any strategy that exploits relative mispricing between different strikes will produce more realistic P&L under SABR than under any other model, because SABR generates the most accurate strike-dependent IV behavior.
Why Binomial Tree Pricing for Tier 3
The Binomial Tree model, introduced by Cox, Ross, and Rubinstein (CRR) in 1979, takes a fundamentally different approach to option pricing.[11]
This step-by-step approach makes the Binomial Tree the most flexible pricer — it makes no assumptions about the volatility process and can price American options accurately regardless of the underlying price dynamics. This flexibility comes at a computational cost (it is slower than Black-Scholes or Bjerksund-Stensland), but for Tier 3 models where volatility dynamics are complex (GARCH clustering, Bates jumps+SV, SABR smile), this flexibility is essential for accurate Greeks.
What each model captures — from constant volatility (GBM) to the full volatility smile (SABR) Choosing the Right Model for Your Strategy
The following decision guide maps common options strategies to their most informative model:
Strategy Type Key Risk Factor Recommended Model Why
Covered calls, cash-secured puts Directional risk GBM or Heston Start simple; add IV dynamics with Heston Vertical spreads (bull/bear) Directional + spread width Heston Skew dynamics affect wing pricing Iron condors, iron butterflies IV level + sudden moves Bates Test against both gradual vol changes and sudden jumps Straddles, strangles Volatility expansion GARCH Calm-to-storm regime transitions are the key risk Calendar spreads, diagonals Term structure of IV Heston or SABR Time-dependent volatility affects front vs back month differently Butterflies, ratio spreads IV smile shape SABR Strike-dependent IV is the core driver of P&L Directional options (long calls/puts) Trend persistence FBM Test whether trend continues or reverts Earnings plays, event-driven Gap risk Jump Diffusion Isolates the impact of sudden price gaps Stress testing any strategy Worst-case scenario Bates + Market Crash The most extreme realistic conditions in the simulator
💡 Progressive testing workflow: Start with GBM (does the basic math work?), then Heston (does it survive realistic IV?), then Bates + Market Crash (does it survive the worst?). If your strategy passes all three, you have strong evidence of robustness.What This Means for Your Trading
Understanding the model behind your simulation is not an academic exercise — it directly determines the value of your test results:
A strategy that only works under GBM is not validated. GBM's constant volatility assumption flatters strategies that would struggle with real-world IV dynamics. Always re-test under Heston at minimum.Match the model to the risk you're testing. If you're worried about gap risk, use Jump Diffusion or Bates. If you're worried about IV crush, use Heston. If you're testing a butterfly's sensitivity to skew, use SABR. Testing the wrong risk with the wrong model produces false confidence.The tier system reflects real complexity. Essential models are faster and simpler — good for learning mechanics. Professional Grade models are slower but more realistic — necessary for serious strategy validation. This is the same trade-off that quantitative trading desks navigate daily.The pricer matters as much as the generator. An option chain calculated with Black-Scholes will look different from one calculated with Binomial Tree, even on the same price chart. The pairings in Options Simulator are designed to maintain mathematical consistency — don't second-guess them without understanding why they exist.
Explore All 7 Models in the Simulator
Frequently Asked Questions
Why does Options Simulator use different option pricers for different models?
Each option pricer has assumptions that must be consistent with the price generator. Black-Scholes assumes constant volatility, making it consistent with GBM and Jump Diffusion. Bjerksund-Stensland handles American options under non-constant volatility, matching Heston and FBM. The Binomial Tree makes no volatility assumptions and handles early exercise, making it suitable for the most complex models (GARCH, Bates, SABR).
Which model is most realistic for U.S. equity options?
Heston or Bates. U.S. equity markets exhibit strong negative correlation between price and volatility (the leverage effect), persistent volatility clustering, and occasional jumps from earnings or macro events. Heston captures the first two; Bates captures all three. Both are widely used by practitioners for equity options pricing and calibration.
Is GBM ever useful, or is it always too simplistic?
GBM is genuinely useful as a baseline and for learning. It isolates the fundamental options mechanics — Delta, Gamma, Theta behavior, time decay acceleration, moneyness effects — without confounding variables. Think of it as a controlled experiment: test your strategy under ideal conditions first, then add complexity. If a strategy fails under GBM, adding stochastic volatility or jumps will not save it.
What is the Hurst exponent in the FBM model?
The Hurst exponent (H) controls long-range dependence. At H = 0.5, FBM equals standard Brownian motion (no memory). At H > 0.5 (e.g., 0.7), price movements show positive persistence — trends tend to continue. At H < 0.5 (e.g., 0.3), movements tend to reverse — mean-reversion dominates. Financial markets typically show H slightly above 0.5 over medium horizons, suggesting weak trend persistence.
Can I change which option pricer is used for a given model?
In Options Simulator, the model-pricer pairings are configured in the admin panel and reflect mathematically consistent defaults. The pairings are: GBM/Jump Diffusion → Black-Scholes; Heston/FBM → Bjerksund-Stensland; GARCH/Bates/SABR → Binomial Tree. Changing these pairings is technically possible but may produce inconsistent or unreliable options pricing.
References & Sources
Black, F., Scholes, M. (1973). "The Pricing of Options and Corporate Liabilities."
Journal of Political Economy , 81(3), 637-654.
DOI
Merton, R.C. (1976). "Option Pricing When Underlying Stock Returns Are Discontinuous."
Journal of Financial Economics , 3(1-2), 125-144.
DOI
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
Desmettre, S., Korn, R., Sayer, T. (2015). "Option Pricing in Practice — Heston's Stochastic Volatility Model."
In: Mathematics of Finance , Springer, 223-270.
Link
Mandelbrot, B.B., Van Ness, J.W. (1968). "Fractional Brownian Motions, Fractional Noises and Applications."
SIAM Review , 10(4), 422-437.
DOI
Bjerksund, P., Stensland, G. (2002). "Closed Form Valuation of American Options."
Working Paper, Norwegian School of Economics.
DOI
Engle, R.F. (1982). "Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation."
Econometrica , 50(4), 987-1007.
DOI
Bollerslev, T. (1986). "Generalized Autoregressive Conditional Heteroskedasticity."
Journal of Econometrics , 31(3), 307-327.
DOI
Bates, D.S. (1996). "Jumps and Stochastic Volatility: Exchange Rate Processes Implicit in Deutsche Mark Options."
Review of Financial Studies , 9(1), 69-107.
DOI
Hagan, P.S., Kumar, D., Lesniewski, A.S., Woodward, D.E. (2002). "Managing Smile Risk."
Wilmott Magazine , September, 84-108.
Link
Cox, J.C., Ross, S.A., Rubinstein, M. (1979). "Option Pricing: A Simplified Approach."
Journal of Financial Economics , 7(3), 229-263.
DOI
Practice This Strategy
Ready to test your strategy across all seven models? Start with GBM, progress through Heston, and stress-test with Bates — all with virtual money and real-time Greeks.
Open Simulator
What to Read Next
Previous: How Simulated Data Extends Options Backtesting
Related: Backtesting vs. Forward Testing on Simulated Data
Find your strategy: Strategy Selection by Market Conditions →
