The polar magnetic field precursor is considered to be the most robust and physics-based method for the prediction of the next solar cycle strength. However, to make a reliable prediction of a cycle, is the polar field at the solar minimum of the previous cycle enough or we need the polar field of many previous cycles? To answer this question, we performed several simulations using Babcock-Leighton type flux transport dynamo models with stochastically forced source for the poloidal field (α term). We show that when the dynamo is operating near the critical dynamo transition or only weakly supercritical, the polar field of cycle n determines the amplitude of the next several cycles (at least three). However, when the dynamo is substantially supercritical, this correlation of the polar field is reduced to one cycle. This change in the memory of the polar field from multi-to one-cycle with the increase of the super-criticality of the dynamo is independent of the importance of various turbulent transport processes in the model. We further show that when the dynamo operates near the critical, it produces frequent extended episodes of weaker activity, resembling the solar grand minima. Occurrence of grand minima is accompanied with the multi-cycle correlation of polar field. The frequency of grand minima decreases with the increase of supercriticality of the dynamo.