Senior Game Math Engineer

• Use at least 1,000,000 spins for directional checks and 20,000,000+ for sign-off.
• Report standard error and confidence interval; reject sign-off if drift exceeds tolerance.
• Keep seeds reproducible and preserve run configuration for replay.

5. Validate generated artifacts.

• Verify book weights are positive, normalized, and mapped to valid state/outcome payloads.
• Recompute empirical RTP/hit-rate/volatility from generated books, not only from formula sheets.
• Confirm replay/event outputs do not mutate payout totals post-generation.

6. Prepare sign-off handoff.

• Deliver assumptions, math decomposition, simulation method, results table, and open risks.
• Include implementation deltas by file path and exact verification commands.

Project Commands

Use existing project verifiers first:

python3 Engine/scripts/verify_rtp.py <game_id> --spins 1000000
bash games/Darumas/verify_math.sh
bash games/Darumas/test_rtp_check.sh

When game-specific scripts differ, keep command shape the same and report the substituted paths.

Output Contract

When designing or auditing math, return:

1. Math Spec: assumptions, mode definitions, EV decomposition, target metrics.
2. Results: theoretical vs simulated RTP/hit-rate/volatility with deltas and pass/fail.
3. Patch Plan: exact files/functions to edit and why.
4. Verification: runnable commands and expected pass criteria.
5. Risks: unresolved constraints that block sign-off.

References

• references/workflow.md: detailed step-by-step execution order.
• references/metrics-and-thresholds.md: formulas, tolerances, and acceptance gates.
• references/signoff-template.md: final report template for handoff.

Execution Rules

• Distinguish theoretical RTP from simulated RTP in every report.
• Treat sub-million-spin conclusions as preliminary only.
• Flag contradictions between max-win marketing claims and math reality as blockers.
• Prefer deterministic, replayable evidence over narrative claims.