A statistical approach to CFD code certification is proposed, which combines verification and validation of CFD simulations with extensions of N-version testing as basis for establishing probabilistic confidence intervals for CFD codes for specific benchmark applications and certification variables. Code modeling and numerical uncertainties are further decomposed into bias and precision uncertainties. Verification and validation provides simulation level bias uncertainty and N-version testing provides code level precision uncertainty. Nversion testing extended for consideration bias uncertainties and use of reference values (experimental data and uncertainties) for estimating interval of certification. Certification is similar to validation, but with additional confidence due to inclusion code precision uncertainty. An example is provided for ship hydrodynamics CFD codes based on the Gothenburg 2000 Workshop on CFD in Ship Hydrodynamics.