A length-based model has been used to estimate stock abundance of Bristol Bay red king crabs (Paralithodes camtschaticus) since the model was developed in 1993. In this study, we conducted a retrospective analysis of the performance of this model for assessment and fishery management. Results from the model assessment in 2000 were used as baseline reference estimates. Retrospective and reference estimates have common trends over time, which are similar to survey estimates. Assumptions about natural mortality and indirect fishing mortality affect recruitment and abundance estimates. Due to low natural mortality and indirect fishing mortality during the 1990s, retrospective estimates tend to be biased low for mature female and large male crabs and biased high for small male crabs. Relative errors derived from differences between retrospective and reference estimates for Bristol Bay red king crabs range from-20% to 11%, much smaller than those derived from major groundfish stock assessments in the eastern Bering Sea. Stock-recruitment curves estimated in terminal years 1993-2000 generally have a similar shape. Although historical errors in stock assessments may not affect selection of the current harvest strategy, their impact on annual guideline harvest level can be substantial in some years when they trigger a shift in harvest rate because the target harvest rate is a step function of stock size.