Skip to main content

Calculations run on this device. Scenario values are not sent anywhere — the engine is fully local.

ai calculator · free

Retrieval Threshold Tuner

Slide the similarity threshold across a scored corpus and trace the F1 curve.

Recall at 90% — what does the precision cost?

Engine 1.0.0 · calculation
0.71
Kept candidates
10 / 15
True in corpus
8
0.50
F1 0.73
0.55
F1 0.73
0.60
F1 0.76
0.65
F1 0.70
0.70
F1 0.78
0.75
F1 0.75
0.80
F1 0.57
0.85
F1 0.50
0.90
F1 0.40

Blue bar = your threshold · green = best F1 (0.70).

Precision
70.0%7 kept-relevant of 10 kept
Recall
87.5%7 of 8 relevant found
F1
0.778
Verdict
Near the best trade

Method

  • Recomputes precision, recall, and F1 at each threshold from a baked scored list with true labels, and precomputes the F1 curve.
  • High threshold buys precision and pays recall; the green bar marks the empirical best trade on this corpus.