This lesson's interactive features are locked, please to keep using them
F1 Score
Precision and recall often pull in opposite directions. So, how do you optimize for both simultaneously?
Enter the F1 score – a single metric that balances precision and recall when both are equally important.
F1 score is the harmonic mean of precision and recall. It gives you one number that represents the overall performance of your search system.
f1 = 2 * (precision * recall) / (precision + recall)
The "harmonic" mean is often better than a "regular" mean (average) because it punishes extreme disparities more:
| Precision | Recall | Mean | F1 Score |
|---|---|---|---|
| 100% | 100% | 100% | 100% |
| 50% | 50% | 50% | 50% |
| 90% | 30% | 60% | 45% |
| 75% | 25% | 50% | 37.5% |
| 100% | 0% | 50% | 0% |
F1 encourages you to improve both metrics! A system that's great at precision but terrible at recall gets a lower F1 score. F1 is perfect when:
- Precision and recall are equally important
- You want a single metric to optimize
- You're comparing different systems
However, if one metric matters more than the other, consider weighted F-scores or just focus on the more important metric directly.
Assignment
Update the evaluation script to also print F1 scores for each test case.
-
k=4 - Query: terrifying bear forest attack - Precision@4: 0.7500 - Recall@4: 0.5000 - F1 Score: 0.6000 - Retrieved: Grizzly, Claws, Prophecy, Day of the Animals - Relevant: Grizzly, Bear, Into the Grizzly Maze, Claws, Unnatural, Day of the Animals - Query: true wildlife bear documentary - Precision@4: 0.2500 - Recall@4: 0.5000 - F1 Score: 0.3333 - Retrieved: Grizzly Man, Grizzly, Prophecy, Ultime grida dalla savana
Run and submit the CLI tests.