1. Numerical Methods for Option Pricing
Prof: Olivier Pironneau
Kimiya Minoukadeh
Ecole Polytechnique
M2 Mathématiques Appliquées, OJME

2. Agenda Introduction to Monte-Carlo method
Heston stochastic volatility model using M-C
Basket option using Monte-Carlo
Accuracy of Monte-Carlo methods
Variance Reduction methods
Conclusion

3. Monte-Carlo Method I
Based on the expectation of a random variable X, given
N samples {X1,X2,…,XN}
Price of a European Call option is therefore calculated as
where: is the ith estimate of the stock price at time T, the time of maturation, r is the risk free interest rate and K is the strike price.

4. Monte-Carlo Method II
The stock price St follows the stochastic differential
Equation (SDE)
where
is the drift term
is the volatility

5. Agenda Introduction to Monte-Carlo method
Heston stochastic volatility model using M-C
Basket option using Monte-Carlo
Accuracy of Monte-Carlo methods
Variance Reduction methods
Conclusion

6. Heston Stochastic Volatility I
Studies have shown that the volatility , if held constant, does not reproduce observed market data. We therefore consider the model suggested by Heston
volatility of stock
rate of mean reversion
volatility mean
volatility of volatility
The cost of the call at time t = 0

7. Heston stochastic volatility II
Results are consistent with the a priori lower bounds known for call options.

8. Heston stochastic volatility III
Barrier options pose the constraint that a certain asset is never allowed to reach outside a certain interval [a,b].
Expectation of payoff considerably reduced
Price of option reduced
a = 0
b = 130

9. Agenda Introduction to Monte-Carlo method
Heston stochastic volatility model using M-C
Basket option using Monte-Carlo
Accuracy of Monte-Carlo methods
Variance Reduction methods
Conclusion

10. Basket options I
Sometimes a derivative may be based on more than one underlying asset. {S(1),S(2),…,S(p)}
The Black-Scholes equation becomes p-dimensional. We consider the case of two underlying assets: p = 2, and once again the Brownian motions have a correlation
Payoff is based on the sum of the two stocks at time T

11. Basket options II Suppose we use
L starting prices of each of the two stocks
N samples of the estimated stock prices
M intervals for the calculations of the stock prices using explicit Euler’s method
Complexity of the program would be O(L2NM).
To reduce this by a factor M to O(L2N), we use Ito’s Lemma with Yi = log(S(i)) to obtain the explicit solution to the SDE
I will now talk about how we implemented the diffusion term into the equation.

12. Basket options III
By using the explicit solution we can observe that we get desirable results, accuracy similar to using the Explicit Euler’s method, however time performance improved dramatically.
TIME PERFORMANCE
ERROR ANALYSIS

13. Basket options IV
Letting K = K1 + K2 for the respective quasi strike prices of stocks S(1) and S(2), we observe the following results
By choosing S0(1) = K1 = 100, we observe that results resemble that of a standard European call option with one underlying asset

14. Agenda Introduction to Monte-Carlo method
Heston stochastic volatility model using M-C
Basket option using Monte-Carlo
Accuracy of Monte-Carlo methods
Variance Reduction methods
Conclusion

15. Accuracy of Monte-Carlo method
The central limit theorem shows that the accuracy of the Monte-Carlo method is controlled by
Thus to halve the error we would need to quadruple the number of samples N used in the Monte-Carlo simulation.

16. Agenda Introduction to Monte-Carlo method
Heston stochastic volatility model using M-C
Basket option using Monte-Carlo
Accuracy of Monte-Carlo methods
Variance Reduction methods
Conclusion

17. Variance Reduction Methods I
IDEA: Reduce the variance of the random process X.
For an independent random process Y, we note that
The variance is then given by
therefore we have
I will now talk about how we implemented the diffusion term into the equation.

18. Variance Reduction Methods II
Need to choose a random variable Y such that it is closely correlated with X.
We adapt a method suggested by P. Pellizzari [1] for variance reduction of basket options
I will now talk about how we implemented the diffusion term into the equation.
[1] P. Pellizzari. Efficient Monte-Carlo pricing of basket options. Finance, EconWPA, 1998

19. Variance Reduction Methods III
We see that we considerably improve the accuracy of the Monte-Carlo method when using variance reduction technique.
With variance reduction, we obtain with N = 2000 samples, results as accurate as the normal Monte-Carlo method with N=10000 samples.

20. Agenda Introduction to Monte-Carlo method
Heston stochastic volatility model using M-C
Basket option using Monte-Carlo
Accuracy of Monte-Carlo methods
Variance Reduction methods
Conclusion
To demonstrate different methods for solving ODEs we will consider a simple cell model based on the Fitzhugh Nagumo model, which as described before is a simplification to the Hodgkin-Huxley model. For v, we have chosen the initial condition 0.2 which is above the threshold for the action potential to take place. The graph to the top right shows the result.

21. Conclusion
The Monte-Carlo method is intuitive and extremely easy to implement
It can be used to calculate call prices when an analytic solution of a PDE does not exist
Data is consistent with observed data
For well estimated expectations we need many sample simulations. To double accuracy, number of samples must quadruple.
IMPROVEMENT: When analytic solutions do not exist and we are obliged to use Monte-Carlo methods, variance reduction can improve the performance of the calculation.