overfit

Today it means: To use a statistical model that has too many parameters relative to the size of the sample, leading to a good fit with the sample data but a poor fit with new data. Full definition →