Two new stopping criteria designed to terminate automatically inactive cycle runs in Monte Carlo (MC) neutronics eigenvalue calculations are derived in terms of the covariance matrix of stochastic error components inherent in the stationary fission source distribution (FSD) from each stationary MC cycle run. A practical way to determine the covariance matrix using nonstationary FSD in the course of inactive cycle MC runs is presented. The effectiveness of the new stopping criteria including the way to calculate the covariance matrix is examined through continuous energy MC neutronics calculations for ten pressurized water reactor test problems with varying dominance ratios. It is shown that the empiricism-free new stopping criteria stop inactive cycle MC runs effectively and that FSDs from the termination of inactive MC runs are stationary consistent with the available posterior source convergence diagnosis.