Public scorecard
The forecast record, audited.
Every watch item resolves. Every resolution carries a directional read assessment, a horizon call, and a significance check. They tally here. Aligned, mixed, misread — every directional call counts. Calibration is the moat; the discipline is reading the field accurately enough that the observatory's reads can be trusted next cycle.
2 resolved · 6 pending
Directional accuracy0%0 aligned · 0 mixed · 2 misread
Horizon accuracy50%1 within · 0 early · 1 late
Significance accuracy50%1 confirmed · 1 overstated · 0 understated
Last 2 resolutions
Per-place calibration
How the score is computed
- Directional accuracy = (aligned + 0.5 × mixed) / total resolved. Mixed reads receive half-credit because they capture partial-truth dynamics. Misreads and unforeseen outcomes earn zero.
- Horizon accuracy= within / total resolved. Whether the predicted event happened inside the watch's stated horizon window. Early and late are both not-within.
- Significance accuracy = confirmed / total resolved. Whether the predicted cardinal weight matched the actual structural impact of the outcome.
- Misreads are reported. The point is calibration, not marketing. Every resolved watch shows its assessment publicly, including those where the observatory was wrong.