The free-agency class, priced as risk

Risk is the odds of keeping less than 80% of platform-year portable production over the two seasons after signing. Bands anchor at measured 2019-2023 rung-by-spike rates (incumbent 23.5%, external 43.5%; with a career-year spike 36.7% and 56.3%) and the age-cliff flag escalates one band. Expected retention comes from a frozen regression on the same frame; portability flags are advisory until a larger sample prices them. Letting him walk has a price too: walkCompRound projects the compensatory selection (two drafts out) his departure would earn at roughly his expiring cap share, gross of the net-loss formula - qualifying signings cancel losses, so the realized pick depends on the rest of the signing ledger. Odds, not verdicts: the incumbent ladder is the best public sensor for what stats cannot see, and it is probabilistic - players walk and win DPOY. Guarantee share is the expiring deal structure; dead-weight risk concentrates where heavy guarantees meet career-year spikes.

These forecasts are sealed and graded in public on the prediction ledger.

Likely rooms are external destination odds for market setters only, where a market setter is positional, not a flat dollar bar: an expiring deal one standard deviation above the position's mean, or a projected next deal that clears the same bar. The projection runs on platform production and age, plus the coverage counting line for corners and safeties (held-out correlation .74 with realized deals), so rookie-contract risers gate in and faded names gate down; together the two doors caught 69% of the players who actually set markets on held-out classes, against 46% for the expiring deal alone. Proj market is the projected next deal as a share of the cap, converted at the class year's implied cap; on held-out paid signings it missed by about $6.6M per year on average and it still under-calls the very top of the market, so read it as a floor-ish estimate, not a sticker. The destination model itself runs on roster state (need, cap commitment, losing seasons); under the expiring gate its top three named the actual signer about 20% of the time on held-out 2025 and 2026 signings against 10% blind, and the projected door's own destination hit rate is unmeasured until the March ledger grades it. Odds are provisional until the need boards refresh, they exclude the incumbent because re-signing is a different market, and a decline tag marks players whose age cliff or severe external-buy risk says the expiring deal may overstate the market itself.