Parallel Binomial American Option Pricing with (and without) Transaction Costs

TL;DR

This paper introduces a parallel algorithm for American option pricing with transaction costs using binomial trees, optimized for multi-core processors, demonstrating significant speedup and efficiency improvements.

Contribution

It presents the first parallel binomial tree algorithm that incorporates transaction costs, with dynamic load balancing and synchronization for modern multi-core systems.

Findings

Parallel speedup of 5.26 with 8 processors

Parallel efficiency of 65.75% achieved

Algorithm effectively handles transaction costs in option pricing

Abstract

We present a parallel algorithm that computes the ask and bid prices of an American option when proportional transaction costs apply to the trading of the underlying asset. The algorithm computes the prices on recombining binomial trees, and is designed for modern multi-core processors. Although parallel option pricing has been well studied, none of the existing approaches takes transaction costs into consideration. The algorithm that we propose partitions a binomial tree into blocks. In any round of computation a block is further partitioned into regions which are assigned to distinct processors. To minimise load imbalance the assignment of nodes to processors is dynamically adjusted before each new round starts. Synchronisation is required both within a round and between two successive rounds. The parallel speedup of the algorithm is proportional to the number of processors used. The parallel algorithm was implemented in C/C++ via POSIX Threads, and was tested on a machine with 8 processors. In the pricing of an American put option, the parallel speedup against an efficient sequential implementation was 5.26 using 8 processors and 1500 time steps, achieving a parallel efficiency of 65.75%.