The construction industry has been forever blighted by delay and disruption. To address this problem, this study proposes the Fitzsimmons Method (FM method) to improve the scheduling performance of activities on the Critical Path before the project execution. The proposed FM method integrates Bayesian Networks to estimate the conditional probability of activity delay given its predecessor and Support Vector Machines to estimate the time delay. The FM method was trained on 302 completed infrastructure construction projects and validated on a ?40 million completed road construction project. Compared with traditional Monte Carlo Simulation results, the proposed FM method is 52% more accurate in predicting the projects' time delay. The proposed FM method contributes to leveraging the vast quantities of data available to improve the estimation of time risk on infrastructure and construction projects.