The Competitive Prize Collection Problem (CPCP) involves two autonomous players competing to collect a fixed number of contestable prizes located in the Euclidean plane. We propose a Strategic Planning Architecture (SPA) for hierarchically structuring the de-cision problem of each player in terms of planning horizon, aggregation of prizes, contin-gency planning, cognizance of opponent, and dynamic response monitoring. To evaluate the difficulty of problem instances, and compare the performance of player strategies, we design a computational tournament. However, the effectiveness of a player strategy for the CPCP cannot be evaluated in isolation from other strategies, since a strategy can only be evaluated against another strategy, one for each player. In addition, to address the circular nature of benchmark problem instances and effective strategies, we consider first robustness of performance on a general class of problem instances followed by expected performance on a class of bad-case problem instances..