How to interpret residual plot from a regression model

Residual is the difference between the actual data point’s value and the predicted value. Since regression is about predicting continuous set of values, residual scale is also not limited. However, an unbiased model should have residuals distribution symmetric about 0. If residual plot on the training set is more or less symmetric, but that on the test set is not symmetric, then definitely the model is biased, and perhaps overfitting on the training set.