Kirk Borne's answer to What kind of data scientist do other data scientists hate? - Quora

Not doing any cross-validation on the model, or else doing validation with the same data (training data) as used to train the model. These practices can lead to overfitting. This practice can lead to underfitting and/or bias in the models. Declaring that their "answer" or "solution" is known with absolute certainty. Implying causation when only correlation has been demonstrated.