Calibration

Your confidence record

The lab tracks whether confidence fits evidence strength. A high-confidence wrong call is treated as a different signal from a low-confidence miss.

Local calibration

Judgment signal

Submitted0
Score0
Calibration0
High-conf misses0

Scores stay in this browser. The profile gets useful after several completed cases.

How To Read This Page

The calibration score is a mirror, not a verdict. High scores suggest that your confidence was roughly aligned with the evidence in completed cases. High-confidence misses are the most useful learning events because they reveal where a tidy story felt stronger than the evidence.

Low confidence is not failure. In many data science settings, the most responsible answer is a bounded recommendation with explicit uncertainty. The goal is to become appropriately confident, not permanently cautious.

What To Do After A Miss

  • Reopen the case replay and compare your cited evidence with the expert reconstruction.
  • Identify whether the miss came from a metric definition, causal leap, model-use assumption, or operational constraint.
  • Write the sentence you would use in a real meeting to hedge the claim without becoming vague.