Precision & Recall

Precision asks how many flagged items were correct; recall asks how many relevant items were found. When you change the decision threshold for one fixed scoring model, improving recall often lowers precision and vice versa. A better model can improve both. F1 is their harmonic mean.