How can the dual martingale help solving the primal optimal stopping problem?

Battle Count: 123

Highly relevant for quantitative finance practitioners dealing with Bermudan/American option pricing and hedging. The variance reduction technique (factor 25-1600) significantly improves Monte Carlo pricing accuracy, which is critical for real-time risk management, derivative desk pricing, and hedging. However, it is more of a computational/numerical method than a trading strategy. The dual martingale framework also has direct hedging applications as noted in the paper.

Moderate to high complexity. Requires implementing: (1) the pure dual algorithm from Alfonsi et al. 2025 to compute M̂ (involving backward least-squares optimization over martingale increments), (2) the Longstaff-Schwartz algorithm with polynomial regression, (3) the control variate estimator with optimal λ computation, and (4) careful management of three independent sample sets to avoid bias/overfitting. The mathematical framework is well-specified but requires solid understanding of martingale theory and stochastic calculus.

The paper provides detailed algorithmic descriptions (Appendix B), explicit parameter settings for all numerical experiments (Q1, N̄, P, polynomial order, number of samples), and references to the prior work (Alfonsi et al. 2025) for the dual martingale construction. However, no code repository is explicitly mentioned. The methodology is fully specified mathematically with equations (7), (8), (9), (14), (20), (21), (22).