I am working with unbalanced classes to solve a classification problem (whether individuals pay their fees or not). My class imbalance is 75% positive (paid) and 25% negative (unpaid). I have read that stratified kfolds is useful because it preserves the class balances while computing each fold and performing cross validation. However, I found stratified kfolds to perform worse than regular folds for my data. Can this happen or does something seem wrong? Unfortunately I cannot share the data and I realize the explanation might not be tailored to the specific data I have, but what are potential reasons for stratified kfolds performing worse?