What is prediction tracking?

Prediction tracking is a concept in personal epistemology: Recording what you expect to happen and comparing to what actually happens is the only reliable method for calibrating judgment. Without a written record, hindsight bias rewrites your memory of what you believed, making genuine learning from experience impossible.

Example: A product manager believes her company will close a major enterprise deal within six weeks. She is 85% confident. Rather than holding this prediction in her head, she opens her prediction journal and writes: "2026-02-22 — Enterprise deal with Acme Corp closes by April 5. Confidence: 85%. Reasoning: Champion is enthusiastic, budget is approved, legal review is underway. Key risk: Their CTO has not signed off." Eight weeks later, the deal closes — two weeks late, after the CTO demanded a security audit that delayed the contract. She scores the prediction: the event happened, but outside her timeframe. Her 85% confidence was overconfident given that she had flagged but underweighted the CTO risk. She adjusts: in future deals, unresolved executive stakeholder concerns get a 15-point confidence reduction. Without the journal entry, she would have remembered herself as having predicted the deal would close — which it did — and learned nothing about her systematic underweighting of stakeholder risk. The record made the miscalibration visible. Visibility made correction possible.

This concept is part of Phase 8 (Perceptual Calibration) in the How to Think curriculum, which builds the epistemic infrastructure for perceptual calibration.