Models

The four seated models (v6n · v6n2 · v9n · modelV31) plus two shadow challengers, each judged on its own top-5 desk: equal-weight open→close, net of 40 bps per session. Only the four seats vote in The Book; shadow AM histories are current-corpus reconstructions, not live-forward records.

Quantsheets — full model tearsheets & verdict board · design-vs-OOS accrual for every registered book, plus the production-tracked candidates below.

Cards show each model's own top-5 (by |score|), independent of the consensus gates. AM rows are canonical (bucket closed 06:15 PT, entry at that session's open, settled at its close). PM preview is the 18:00 PT snapshot of the still-forming bucket — picks can change by morning; previews are kept frozen and settled later so you can audit them. Seat stats include the replay window from 2026-04-27; shadow displays are date-filtered from the 2026-07-18 re-add boundary (first eligible session 2026-07-20) but reconstructed on each AM refresh. The muted July figures are replay context only. Tradeability floor (added 2026-07-26, applied to the full window): picks require a prior-session close ≥ $1 — sub-dollar names are unborrowable/unfillable at size and their prints unreliable (a stale 4-cent quote once booked a fictional −161% 'short loss'). Click a model for its full history, equity curve and methodology.

Paper/forward accrual only — each track is judged at its own pre-registered clock, not before. Full tearsheets and the verdict board: Quantsheets.

Production-tracked candidates (operator 2026-07-19). Paper/forward accrual only; no capital. Judged at each track's pre-registered clock.

v6n attention surprise — share chassis, human-author gates

v6n2 two-flow surprise — shorts only from surging bear flow

v9n commitment-only two-flow (CD+PD, feeds excluded)

modelV31 credible-breadth consensus — the diversifier seat

v6b EWMA-baseline surprise — reconstructed shadow challenger

v9 commitment-only surprise (CD+PD) — reconstructed shadow challenger