Forecast Accuracy
At YearlyFuture we keep records, not promises. The moment a prediction is published, its verification criterion, probability and sources are frozen; its outcome is published on this page.
Tracking Status
- Published and tracked: 160
Forecast Accuracy
- Resolved: 0 (correct: 0 · incorrect: 0 · unverifiable: 0)
- Hit rate: no result yet
- Brier score: no result yet
- Average probability: 58.2% · Recorded probability updates: 28
Probability Calibration
| Probability band | Predictions | Our stated average | Actual outcome |
|---|---|---|---|
| %0-19 | 14 | 5% | awaiting data |
| %20-39 | 8 | 32% | awaiting data |
| %40-59 | 52 | 53% | awaiting data |
| %60-79 | 62 | 67% | awaiting data |
| %80-99 | 24 | 87% | awaiting data |
Why are there 0 results today? (and why that is good news)
Our first predictions target 2027-2030. A prediction cannot be resolved before its target year is complete. That is why you see 0 results today: we do not manufacture a score afterwards, we accumulate it over time. A score that can be adjusted later is not a score.
Most changed predictions
- 2027 Sonuna Kadar, Önde Gelen Bir Yapay Zeka Laboratuvarı, FrontierMath Benchmark'ının En +25
- Kuantum Bilgisayarlar 2033'e Kadar İlk Kez Ticari Bir Endüstri Probleminde (Malzeme Tasarı +20
- AB'nin Yüksek Riskli AI Sistemleri İçin Ağustos 2026'dan Aralık 2027'ye Ertelenen Uyum Tar +15
- 2027 Sonuna Kadar, Küresel Ölçekte AI Sunucularının Elektrik Tüketimi, Geleneksel (AI-dışı +15
- Moderna Ve Merck'in Kişiselleştirilmiş MRNA Kanser Aşısı (İntismeran İle Keytruda Kombinas +15
What you can check today
- Every published prediction carries a verification criterion written on the day it was published; it cannot be changed later.
- Every prediction has its probability and full change history on record.
- The sources behind each prediction are listed openly on the prediction page.
- Probabilities are updated as evidence changes; the reason and the date stay in the history.
How outcomes are recorded
An outcome is recorded only with a verified source and at least "medium" confidence; if the evidence is insufficient the prediction remains "unverifiable". No result rests on opinion or guesswork.
Our commitment
The first results will be published in January 2028. From then on we will report how many predictions came true, our hit rate and our Brier score exactly as they are - including results below our expectations. A platform that hides bad outcomes is not a forecasting platform.