A unified approach, which is valid for both standard and nonstandard singularly perturbed systems, is presented to achieve the composite approximation of the full-order linear feedback saddle-point solution. It is proved that such a composite approximation forms an O(epsilon) (near) saddle-point equilibrium of the full-order game, and the resulting value is O(epsilon(2)) over or below the exact value of the full-order game, depending on the given game parameter.