The bearing plates used in plate load test for highway engineering are typically rigid. However, due to limitations in obtaining the accurate distribution of compressive stress at the bottom of the bearing plate, there is often a significant deviation between the measured subgrade resilient modulus and the actual condition. To address this issue, a flexible bearing plate can be used to test the subgrade and obtain a more accurate resilient modulus. In this study, we use variance and degree of mean deviation to quantitatively evaluate the distribution uniformity of compressive stress. To create a rigid-flexible bearing plate that is similar to a flexible bearing plate, we explore the combinatorial design of steel plates and rubber mats. We examine factors such as the thickness (10, 20, and 30 mm) of the steel plate, elastic modulus (5, 10, and 20 MPa) and thickness (10, 20, and 30 mm) of the rubber mat, friction coefficient (μ:0, 0,2, 0.4, 0.6, 0.8, ∞) between the bearing plate and subgrade, and the combined shape characteristics of the rubber mat and steel plate. To reduce friction between the rubber mat and subgrade, we use lubricant, and through our design process, we develop a flexible bearing plate with relatively uniform compressive stress. Our computations show that when μ = 0.05, the variance is 0.0001, and the degree of mean deviation is 0.0780. These results indicate that the distribution uniformity of the compressive stress is very close to the uniform distribution load, which meets the necessary accuracy requirements for engineering applications.