2017 Predictions: Calibration Results
Read the original on Slate Star Codex →
Follows up on
↳ Predictions For 2017 — Prediction/forecast · Jan 2017
Summary
Scott scores his 2017 predictions — the 100+ items marked true/false, a per-confidence-bucket calibration table (62% on 50s up to 100% on 99s), the calibration graph (neither globally over- nor under-confident), and an error analysis: he overcorrected from last year's 70%-underconfidence into slight overconfidence, showed a 'things go more smoothly than they did' bias on personal affairs (diet, savings, travel, housing), and underestimated the 2017 economy. The scoring companion to the annual predictions list.
Why this score
Quality 56 · Solid. A solid forecasting-practice artifact — public, pre-registered, scored, with a genuinely useful per-bucket calibration analysis — but a results-scoring list rather than a developed essay. Solid, a touch above the bare predictions lists for the calibration analysis.
Claude’s paradigm shift 32 · Slight. A recurring calibration-scoring format; minimal novelty. Slight.
Real-world impact 2 · Minor. Models good forecasting practice; a single year's scoring has little material effect. Within-blog influence.