Fisheries assessments require accurate knowledge of stock biomass and trends to improve sustainability; however, the numerous assumptions made when estimating stock biomass generate high uncertainty. We contrasted past assessments of stock biomass with updated reassessments for 230 fisheries worldwide. Compared to the most recent assessments, past estimates of stock biomass were generally over-stated and initial biomass understated, particularly for low value stocks. These biases led to inappropriate recognition of sustainability for many overfished stocks. We estimate that more than three times as many stocks are collapsed (<10% initial biomass) than currently recognised, and found that rising trends reported for overfished stocks were often misleading. The high uncertainty in modelled stock estimates warrants much greater precaution in management decision making. Consequences of uncertainty could be reduced by considering accumulated parameter error, changing environmental conditions, separate models with contrasting assumptions, retrospective comparisons, and collection of fisheries independent data, including surveys from unfished marine reserves.