Metrics
F1 Score
Definition
The harmonic mean of precision and recall, expressed as a single number between 0 and 1.
Why it matters
It is a convenient summary when you need to balance completeness against purity.
In practice
F1 combines precision and recall into one number using the harmonic mean, which stays low if either one is low. A review with 95% precision and 20% recall scores only about 33%.
F1 is useful for comparing models or cutoffs in testing. In discovery, protocols usually set targets for recall rather than F1, because completeness is what the other side and the court tend to ask about.
Formula
F1 = 2 × precision × recall ÷ (precision + recall), which equals 2 × TP ÷ (2 × TP + FP + FN)
Worked example
With 850 true positives, 150 false positives and 100 false negatives, precision is 85.0% and recall is 89.5%. F1 is 2 × 850 ÷ (1,700 + 150 + 100), or about 87.2%.
Illustrative numbers, not from a real matter.
Compare with
- Recall: F1 can look healthy while recall sits below a protocol target, so report recall on its own too.
Work it out
- Precision, Recall & F1 calculator · How accurate are the model or reviewer calls?