Research preprint · Exercise valuation
Discretionary Exercise by Conditional-Density Reweighting of Monte Carlo Paths
Gregory Pelts ·
Abstract
We study a Monte Carlo procedure for discretionary exercise that reweights simulated futures with short-step transition densities. Its principal advantage over continuation-value regression is a general, model-independent exercise interface: the model supplies conditional weights, the contract supplies cashflows, and the exercise engine requires no choice of regression variables or basis functions tied to internal model factors. Prices remain dependent on the specified dynamics. The construction belongs to the stochastic-mesh family and preserves each original path's selected cashflows for final valuation. We develop the conditional-density identity and its numerical implementation, illustrate a surrenderable Asian contract, and compare Black–Scholes and Heston exercise prices with tree and finite-difference references.