This talk presents a new framework for global optimization based on the strategic law of large numbers. For a possibly multimodal continuous function on a bounded rectangular domain, the report first shows that the global optimization problem is equivalent to an optimal sampling strategy formation problem in a two-armed decision model with known distributions. Theoretical analysis reveals that optimal strategies are generally not unique, and that a concrete strategy based on the sign of the partial gradient of the unique solution to a parabolic partial differential equation is asymptotically optimal. Motivated by these results, the report proposes a class of Strategic Monte Carlo Optimization (SMCO) algorithms, which require no PDE solving and make coordinate-wise two-armed decisions based solely on the signs of the partial gradient (or practically the first difference) of the objective function. Under sufficient conditions, the SMCO algorithm converges to a local optimizer from a single starting point, and to a global optimizer as the set of starting points grows. Extensive numerical studies show that the SMCO algorithms perform robustly on challenging landscapes—multimodal, nondifferentiable, and discontinuous. This approach offers a new theoretical perspective and a practical algorithmic tool for black-box global optimization.