#Evidence — Simulation

Seeded Monte Carlo aggregates for hold-and-win-reference, produced by the canonical Go backend. This is the authoritative Draft evidence for the mechanic, because the round is not enumerable (see the Math Spec §12).

Every report here is estimated. Monte Carlo is never presented as exact.

#Required manifest fields

Each report records, or it is not evidence:

• Backend Git SHA and Go version.
• Canonical config path and SHA-256.
• RNG algorithm/version (sha256-trunc64-be/go1-alfg#1).
• Round count, master seed / seed-set ID, shard boundaries, worker count.
• Elapsed time and the exact commands run.

#Required metrics

• Total, base and hold_and_win RTP, each with standard error and a 95% confidence interval.
• Trigger frequency with an interval.
• Conditional feature award: mean, median and percentiles.
• Steps per feature (duration), reset rate, full-grid probability.
• Per-tier prize frequencies (Regular values, Mini, Minor, Major, full-grid Grand).
• Hit frequency and standard deviation.
• Observed maximum, and confirmation that no round exceeded the 500× bound.

Base results and stateful feature results are reported separately.

#Contents

Empty. Populated at plan Task 10.