Bagging works by sampling a smaller set of data and matching the distribution of the greater set of data—the algorithms are then trained on these smaller sets and later fused to form a single decision. Bagging will only help with learners that are capable—learning a bunch of bad models will lead to bad results.