Skip to content
Tutorials Center
My learning

Topics

Lesson 31 of 38 · Applications in Finance and Science

Option Pricing by Monte Carlo

Price a European call by simulating geometric Brownian motion, check against Black-Scholes, then add variance reduction and price an Asian option.

Finance was one of the first fields to rely heavily on Monte Carlo. A European call option gives its holder the right, but not the obligation, to buy an asset at a fixed strike price K on a fixed date T. If the asset price S_T at that date exceeds K, the holder gains S_T − K; otherwise the option expires worthless. The payoff is max(S_T − K, 0).

What is the option worth today? Modern pricing theory says: under a special "risk-neutral" model, the price is the discounted expected payoff. An expectation is exactly what Monte Carlo is good at. This lesson is an educational model of the technique, not financial advice or a trading tool; real markets are messier than the model.

The model: geometric Brownian motion

Under the standard Black-Scholes model with a constant risk-free interest rate r and volatility σ, the asset price at time T is

Text
S_T = S₀ · exp( (r − σ²/2)·T + σ·√T·Z ),   Z ~ Normal(0, 1)

This follows from geometric Brownian motion; the −σ²/2 term makes the expected price grow at exactly the risk-free rate (E[S_T] = S₀·e^{rT}). The option price is then

Text
C = e^{−rT} · E[ max(S_T − K, 0) ]

So: draw many Z, form S_T, compute payoffs, average them, and discount. The same payoff idea gives puts (max(K − S_T, 0)) and more exotic contracts.

Pricing and checking against Black-Scholes

For this simple option there is also an exact formula, the Black-Scholes formula:

Text
d₁ = [ ln(S₀/K) + (r + σ²/2)·T ] / (σ·√T),   d₂ = d₁ − σ·√T
C  = S₀·N(d₁) − K·e^{−rT}·N(d₂)

where N is the standard normal CDF. Since the answer is known, we can test our simulation.

Python European call: Monte Carlo vs Black-Scholes

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The exact value should lie inside the Monte Carlo interval in about 95% of runs; with the seed shown here it can fall just outside, which is expected about once in 20 runs. Change the seed or raise n to see the interval tighten.

Reducing the variance

Two ideas from earlier in this chapter apply directly:

  • Antithetic variates: use Z and −Z. The payoff is a monotone function of Z, so pairs are negatively correlated.
  • Control variate: S_T itself has known mean S₀·e^{rT} and is highly correlated with the payoff. Subtract β·(S_T − S₀e^{rT}) with β estimated by regression (see control variates).
Python Variance reduction for the call price

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

All three methods agree on the mean; the last two have smaller spread for the same number of payoff evaluations.

Where Monte Carlo is really needed: an Asian option

An Asian option pays off on the average price over the life of the contract, max(Ā − K, 0). The arithmetic average of lognormal prices has no simple closed-form distribution, so Monte Carlo is the natural tool. We simulate the entire price path in small time steps: S_{t+Δ} = S_t · exp((r − σ²/2)Δ + σ√Δ·Z).

Python Asian call vs European call

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The Asian option costs less because the average of a path moves less than its endpoint.

Cautions

  • The model assumes constant volatility and normal log-returns; real markets show jumps and changing volatility.
  • Monte Carlo pricing of early-exercise options (American) needs extra methods such as least-squares regression; it is beyond this lesson.
  • Always compare with a formula when one exists, and report the standard error with every price.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. Under the risk-neutral model, the option price equals:

    Price = e^(-rT) times the expected payoff under the risk-neutral measure.
  2. 2. Why is the term -sigma^2/2 in the exponent?

    The lognormal mean includes a +sigma^2/2 T term, which the correction cancels.
  3. 3. Which of these is a good control variate for a call payoff?

    S_T is strongly correlated with the payoff and E[S_T] = S0 exp(rT) is known exactly.
  4. 4. Why is Monte Carlo especially useful for an arithmetic-average Asian option?

    The average of lognormal prices is not lognormal, so simulation of paths is used.

Summary

  • Option price = discounted expected payoff; simulate S_T = S₀·exp((r − σ²/2)T + σ√T·Z).
  • Check against Black-Scholes and report standard errors.
  • Antithetic and control variates reduce variance at no extra cost per payoff.
  • Path-dependent contracts such as Asian options need full path simulation, where Monte Carlo shines.

Next: Value at Risk by Simulation