IPCC’s fifth assessment report concluded that there are likely more areas with increases in heavy precipitation than decreases. The report states further that confidence in changes in total precipitation globally is medium after 1950 and low before. This raises the question whether global precipitation extremes are increasing at the cost of light events or if intensification happens throughout the distribution. We present here a non-parametric approach to compare precipitation distributions between two periods. As a case study, we present the results of the application of this methodology to Australia. Our current understanding of global changes in precipitation distributions is also limited by the lack of reliable precipitation observations. Existing gauge based gridded datasets of daily precipitation and satellite based observations contain artefacts and have a short length of record, making them unsuitable to analyse precipitation extremes. As such, we also reveal a brand new global land-based dataset of daily precipitation with coverage from 1950 to 2013 and a 1 degree resolution. The largest limiting factor for the gauge based datasets is a dense and reliable station network. Currently, there are two major data archives of global in situ daily rainfall data, first is Global Historical Station Network (GHCN-Daily) hosted by National Oceanic and Atmospheric Administration (NOAA) and the other by Global Precipitation Climatology Centre (GPCC) part of the Deutsche Wetterdienst (DWD). We combine the two data archives and use automated quality control techniques to create a reliable long term network of raw station data, which we then interpolate using block kriging to create this dataset. We evaluate our dataset by comparing it with existing global and regional products such as, NOAA Climate Prediction Centre (CPC) Global V1.0 and GPCC Full Data Daily Version 1.0. We find that our raw station density is much higher than existing daily and even monthly datasets. To avoid artefacts due to station network variability, we provide multiple versions of our dataset based on various completeness criteria, as well as provide the standard deviation, kriging error and number of stations for each grid cell and timestep to encourage responsible use of our dataset. Despite our efforts, the raw station density could be further improved in India, South America and Africa. Our dataset would allow for more reliable global analyses of precipitation including its extremes and pave the way for better global precipitation observations with lower and more transparent uncertainties.