#Book of Odin v2 — Evidence Manifest, draft-v0.1

DEV ONLY. NOT FOR PRODUCTION. NOT FOR CERTIFICATION. Internal measurement of an unfrozen pack. Nothing here is an external certification claim.

#1. What this manifest binds

Every figure below is a statement about these bytes under this RNG. If either moves, the evidence describes math that is no longer in the repository.

#2. Where the evidence lives

#3. Exact results

No sampling. Full enumeration of the reel-stop state space (20⁵ per reelset) and integer convolutions of the reel windows.

#4. Simulation parameters

#5. Reconciliation

whole game = 39.5138%  base lines            EXACT
           + 30.6726%  FS lines + expansion  EXACT
           + 25.0787%  base-origin H&W       SAMPLED
           +  2.7328%  FS-origin H&W         SAMPLED
           ─────────
           = 97.9979%   target 98.0000%   deviation -0.0021 pp   tolerance ±0.10 pp

GRAND (0.2869%) is inside the two Hold & Win terms, not a fifth addend.

#6. Max-win status

The Hold & Win ceiling has no witness and the reason is recorded rather than left implicit: all fifteen cells would have to draw the 100× coin, (1/102)¹⁵ ≈ 7.4 × 10⁻³¹, before accounting for filling the grid. The feature is a stochastic state machine, so enumeration is impossible and no seed search can exhibit it. The change request permits a documented reason in place of a proof.

#7. Skipped, blocked or absent

#8. Defects found and closed during this work

Recorded because each one would have shipped silently and each is now pinned by a test.

#9. Immutability confirmations