Potential fecundity models of walleye or Alaska pollock Gadus chalcogrammus in the Pacific waters off Hokkaido, Japan, were developed. They were compared using a generalized linear model with using either standard body length (LS) or total body mass (MT) as a main covariate along with Fulton's condition factor (K) and mean diameter of oocytes (DO) as additional potential covariates to account for maternal conditions and maturity stage. The results of model selection showed that MT was a better single predictor of potential fecundity (FP) than LS. The biological importance of K on FP was obscure, because it was statistically significant when used in the predictor with LS (i.e. length‐based model), but not significant when used with MT (i.e. mass‐based model). Meanwhile, DO was statistically significant in both length and mass‐based models, suggesting the importance of downregulation on the number of oocytes with advancing maturation. Among all candidate models, the model with MT and DO in the predictor had the lowest Akaike's information criterion value, suggesting its better predictive power. These newly developed models will improve future comparisons of the potential fecundity within and among stocks by excluding potential biases other than body size.