When training a model, why should you randomly split the rows into separate subsets?

Training a model twice on the same data or different subsets without a proper test set does not inherently lead to better accuracy and can still result in an overfit model if not evaluated on unseen data.

While multiple models can be trained, the primary reason for splitting data is not to train them simultaneously for better performance, but rather to properly evaluate a single model's generalization capability.

Randomly splitting data into distinct training and testing sets allows the model to be evaluated on data it has not previously encountered during its training phase. This evaluation helps assess the model's ability to generalize to new, real-world data and detect overfitting.