A heavy hailstorm developed over Istanbul on July 27, 2017 and inflicted substantial damage on hundreds of buildings and thousands of vehicles. Weather forecast models largely failed to predict this event. In this study, we investigate the performance of the Weather Research and Forecasting (WRF) model in simulating this particular hailstorm through a sensitivity experiment and the development of the supercell using one of the simulations that produce comparatively better performance in reproducing the hail event. A total of 24 sensitivity simulations involving different microphysics (Milbrandt, NSSL2, WDM6, and Goddard), cumulus (New SAS, New Tiedtke, and Kain Fritsch), and planetary boundary layer (PBL) (YSU, and MYNN2) schemes are conducted. The sensitivity simulations produce different rainfall patterns in the innermost domain, and the major differences in the spatial patterns arise because of the cumulus parameterizations, while microphysics schemes seem to influence the magnitude of the precipitation. All simulations reproduce thunderstorms resulting in substantial precipitation over Istanbul, however, only those with Kain Fritsch cumulus scheme are able to generate hail over the city. The simulation with YSU PBL, Kain Fritsch cumulus and Milbrandt microphysics schemes (YKM) produces hail paths that best match the observations. The YKM simulation reproduces a near typical structure of a storm cloud with an anvil shape, an overshooting top reaching about 12 km altitude, and a weak echo region in the front part. The model is able to simulate strong convection supported by the low-level convergence and low-to-mid-tropospheric wind shear that produces an updraft near the center of the storm where warm and moist air rises rapidly upward at velocities up to 28 m/s. The simulation yields 21.5 mm precipitation and 3.9 mm hail during the first 15 min when the storm starts to affect the city center. Our analysis suggests that the WRF model has high potential to realistically simulate the hailstorms if physics options are reasonably selected.