Physical model testing in irregular waves is often used to assess performance, and validate numerical models, of offshore structures and platforms, yet the full phase-resolved experimental conditions including reflections and correct nonlinearity are seldom replicated numerically. This results in uncertainty in numerical model validation, particularly when considering extremes. Herein we assess a time-reversal approach for the accurate numerical replication of experimentally generated wave fields. Fifteen irregular sea states, including steep and breaking cases, are generated in two large laboratories in the UK and replicated numerically using the time-domain nonlinear potential flow code OceanWave3D. Through statistical, temporal, and spatial analysis, the method is demonstrated to be effective at re-creating sea states with a wide range of steepness, bandwidth, and duration, in two very different laboratories. Typical values of the coefficient of determination between experimental and numerical surface elevations, , averaged over all measurement locations exceed 0.9 and are significantly larger than the alternative replication approaches assessed. The resulting numerical model outputs, when used as inputs to hydrodynamic models of offshore systems, will provide an improved level of validation, subsequently reducing uncertainty in both hydrodynamic model performance and platform design. The method is also naturally suited to replicating measured ocean data for conditions that can be approximately modelled as two-dimensional.