Forecast Audit · Live
Forecast Audit Ledger
Every quantitative forecast → realized outcome → hit rate. A Damodaran-style honest ledger.
7 US macro indicators (from sister site)
Sister site economic-forecasting.com publishes a daily 1-month-ahead forecast for 7 US macro indicators. This page will cron-hook publication time and the realized print, then compute MAE / directional hit rate / divergence from consensus.
| Indicator | Model | Cadence | Hit rate |
|---|---|---|---|
| Core PCE | ARIMAX + Transformer ensemble | monthly | awaiting econ-forecasting hook-up |
| NFP (nonfarm payrolls) | TabPFN | monthly | awaiting econ-forecasting hook-up |
| Industrial production | ARIMAX | monthly | awaiting econ-forecasting hook-up |
| Initial jobless claims | Transformer | weekly | awaiting econ-forecasting hook-up |
| 10Y UST yield | Random Forest | daily | awaiting econ-forecasting hook-up |
| 2Y UST yield | Random Forest | daily | awaiting econ-forecasting hook-up |
| 10Y-2Y spread | derived | daily | awaiting econ-forecasting hook-up |
Monthly view tracking (this site)
Once a monthly macro_view is committed, the end-of-month review marks it failed / partial / hit. NAV is not used here — this scores the view itself.
Closed
0
Pending
1
Hit rate
—
(hit + 0.5×partial) / closed
| Month | conf. | View summary | Verdict | |
|---|---|---|---|---|
| 2026-06 | ●●●○○ | Inception 三组并行配置. Base 中性 risk-parity + 30% 长债 + 10% 黄金; | PENDING | 详情 → |
Source: data/decisions/*.yaml (publish:true only). Pending verdicts are excluded from hit-rate to avoid the "no conclusion = miss" misread.
Aggregate statistics
Meaningful only after 6+ months. Sample too small now; aggregate numbers withheld to avoid overfitted narrative.
Verdict criteria documented in methodology.md; the same spec is used in resume appendix and interview talking points.