In non-time-series settings, k-fold cross-validation is one of the most popular cross-validation schemes, however it assumes that data is drawn from an independent and identically distributed stochastic process. This assumption is violated when modeling with time-series data, as time series exhibit serial correlation. When k-fold cross-validation is used on time series, it will lead to information leakage between the folds. In particular, it is possible that information from the training set leaks into the validation set. This leads to an overestimation of the model's performance on unseen data, as it is not actually unseen data anymore due to the leakage.