Quantlab-based thesis on Monte Carlo variance reduction
As part of our ongoing collaboration with academia, Algorithmica recently supported a master’s thesis at KTH that applied Quantlab and Qlang to study variance-reduction techniques in Monte Carlo pricing.
The thesis investigates pricing of an up-and-out barrier call under a standard GBM model with fine time discretisation (1024 time steps). Pseudo-random Monte Carlo is compared with Sobol-based quasi–Monte Carlo, both with and without Brownian-Bridge path construction, using Quantlab’s existing pricing infrastructure.
The results confirm well-known effects in a controlled, production-like environment. Sobol sequences significantly reduce error at low to moderate path counts, while Brownian-Bridge construction improves accuracy for barrier options where early path resolution is critical. Runtime differences were negligible in the tested range, and accuracy advantages diminish as path counts increase.
The work helps clarify where these techniques provide tangible benefits, and where increased computational effort yields limited returns, providing useful input for future design and prioritisation decisions.