Different data samples with different numbers of data points are used in the process of constraining cosmological models. For example, Pantheon has 1048 data points, while we have three data points as distance priors of CMB. In practice, CMB has a greater effect on constrained results, while chi-square values obtained from Pantheon play a more important role in the numerical results of the MCMC process. How can this contradiction be justified? on a side, fewer data points, a small effect on chi-square results, and an important effect on constrained parameters. on the other side, a large number of data points, a big effect on chi-square results, and finally a smaller effect on constrained parameters. How we can balance this paradox?