It is required to determine the mechanical properties of Roller Compacted Concrete Pavement (RCCP) such as Compressive Strength (CS) and Splitting Tensile Strength (STS) for constructing and maintaining the road pavement because of the inevitable uncertainties in the experimental setup. In this study, a reliable hybrid intelligent model based on the integration of Grasshopper Optimisation Algorithm (GOA) and Multivariate Adaptive Regression Splines (MARS), called MARS-GOA, was developed to present new closed-form equations for CS and STS prediction of RCCP. Initially, GOA was developed to serve as a search engine of the proposed algorithm to optimise MARS control parameters by minimising the error of prediction. Statistical metrics indicate that proposed hybrid MARS-GOA outperformed ELM, M5p, and standard MARS for prediction of both the CS (CoD = 0.811, PMARE = 22.146% and U95 = 34.670) and STS (CoD = 0.816, PMARE = 16.192% and U95 = 2.725) of RCCP. The results of reliability analysis, namely Monte-Carlo Simulation (MCS), revealed that TCS, TTS, and uncertainty level of the used dataset play a major role in the determination of the reliability of the RCCP mix design.