Desk
Evidence-driven conviction · every number opens to its receipts
—
loading the book…
The Book — tracked positions
Score = 0–111 influence (evidence-earned). Forecasts are the open ledger questions; conviction is the calibration-weighted blend across forecasters.
| Position | Score | Momentum | P(≥1) 30d | Count q10·50·90 | Conviction | Open |
|---|
Open Questions — the slate
The standing exam. Any forecaster — engine baseline, Claude, or yours — answers the same questions; horizons grade everyone.
Forecaster Leaderboard
Ranked by Brier (lower is better). Skill > 0 beats the base rate; wide intervals mean too little history to trust the point.
| Model | N | Brier | Skill | Hit (95% CI) |
|---|
No-Call Board — what the desk refuses to call
Forecasters that declined for lack of evidence, with their reasons. Restraint is a feature: no number here is fabricated.
Signal Tape — fires & fills
Decision-hook fires (forecast crossed threshold) and recorded outcomes (reality answering), newest first.
Desk boundary: This desk tracks OMNIA subjects and their evidence — it holds no securities data, executes nothing, and is not investment or financial advice. Forecast probabilities are claims that will be graded, never certainties; conviction is a transparent calibration-weighted blend, not a recommendation. Influence never equals human worth, and nothing here authorizes JWLS value.