Extreme sea levels, generated by storm surges and high tides, have the potential to cause coastal flooding and erosion. Global datasets are instrumental for mapping of extreme sea levels and associated societal risks. Harnessing the backward extension of the ERA5 reanalysis, we present a dataset containing timeseries of water levels based on a global hydrodynamic model covering the period 1950-2022. This is an extension of a previously published dataset for 1979-2018 (Muis et al. 2020). Using the extended ERA5 dataset, we calculate daily maxima timeseries, statistical percentiles and estimate extreme sea levels for various return periods globally. Validation shows that there is a good agreement between observed and modelled sea levels, with the level of agreement being very similar to that of the previously published dataset. The extended 73-year dataset allows for a more robust estimation of extremes, often resulting in smaller uncertainties than its 40-year precursor. The present dataset can be used for assessing flood risk, climate variability and climate changes. The background of this dataset is described in the corresponding paper (Aleksandrova et al. 2024, paper currently under review).