NBA Sports Game Scheduling Problem and GA-Based Solver

Abstract

The game scheduling regulations of the National Basketball Association (NBA) are discussed and scheduling for it is formulated as a complex traveling tournament problem and referred to as the NBA scheduling problem (NBASP). Benchmarks for the problem were depicted from previous official schedules and organized as formatted files for researchers. Although, reducing the travel length is the goal of the problem, valuation indices: counts of back-to-back games, four-games-in-five-days, and weekend games were reported in the benchmark. A genetic algorithm (GA)-based method is proposed for solving the NBASP, in which a dedicated encoding scheme and decoding procedure were designed to obtain feasible schedules. A tailored population initialization procedure was used to reduce infeasible initial solutions and cost-estimated heuristic mutation operation was employed to effectively accelerate the generation of superior schedules in the simulated genetic evolution of an optimum solution. The 2014-2015 benchmarks of the NBA schedule was used in testing and results showed a reduction the travel length by up to 14%. Moreover, the non-preferable indices for players were also reduced.