The skill of modern wave models is such that the quality of their forecasts is, to a large degree, determined by errors in the forcing wind field. This work explores the extent to which large-scale systematic biases in modelled waves from a third generation wave model can be attributed to the forcing winds. Three different sets of winds with known global bias characteristics are used to force the WAVEWATCH III model. These winds are based on the Australian Bureau of Meteorology’s ACCESS model output, with different statistical corrections applied. Wave forecasts are verified using satellite altimeter data. It is found that a negative bias in modelled Significant Wave Height (Hs) has its origins primarily in the forcing, however, the reduction of systematic wind biases does not result in universal improvement in modelled Hs. A positive bias is present in the Southern Hemisphere due primarily to an overestimation of high Hs values in the Southern Ocean storm tracks. A positive bias is also present in the east Pacific and East Indian Ocean. This is due both to the over-prediction of waves in the Southern Ocean and lack of swell attenuation in the wave model source terms used. Smaller scale features are apparent, such as a positive bias off the Cape of Good Hope, and a negative bias off Cape Horn. In some situations, internal wave model error has been compensated for by error in the forcing winds.