## Scoring & Win System
- Judges: 5 judges score 0–10 on 5 axes (Visual Cohesion, Prompt Creativity, Lyrical Fit, Technical Quality, Originality). Judge Total per judge: 50.
- Battle Score = average of judge totals (0–50) + Finisher Bonus (+2 if the finisher sequence is clear, timed to bar cue, and passes safety) + Clean Run (+1 if zero safety flags). Cap: 53.
- Fan Favor Bonus: fan votes add 0–1.0 pts. fan_share = votes / max_votes_in_battle. FanFavorBonus = round(fan_share * 1.0, 1). New cap with bonus: 54.
- Season Points: Win 3 / Draw 1 / Loss 0. Performance Boost: +1 season point if Battle Score ≥45. Consistency Boost: +1 season point every 3rd consecutive Clean Run. Fan votes never change W/D/L.
- Tie-breakers (order): head-to-head Battle Score; higher Technical Quality; more Finisher Bonuses; higher Fan Favor Bonus; sudden-death 24h prompt-off (new Sora clip; higher Battle Score wins).
- Disqualifiers: hate speech/targeted harassment; real-world violence; unlicensed audio/visuals; repeat submits after safety block without edits; tampering with voting.
- Playoffs: Top 4 by Season Points; seeds by Season Points then avg Battle Score; playoffs use same scoring, Fan Favor Bonus still max +1.0; finals tie → sudden-death prompt-off.

## Prompt Policy & Stage Requirement
- Creators must supply their own prompts; no auto-generated prompts accepted.
- Required stage tag in every prompt: `SBL_NEXUS_STAGE_V1` (hex-grid floor, floating LED monoliths, rim-lit crowd silhouettes, volumetric mist). Entries without the stage tag are invalid.
- Style templates may be purchased/used as scaffolds but must be customized; direct copy/paste of full templates is not allowed.
- Prohibited: real-world locations/conflicts, gore, hate, or PII.

## Round Rules (Violence Allowance)
- Rounds 1 & 2: Stylized tension only (no explicit violence). Focus on lyrical roast, stage spectacle, and cinematic threat without execution.
- Round 3: Stylized “finishers” allowed; still no gore, dismemberment, or hate content. Violence must remain abstract (light, particles, dissolves, petals) and clearly fictional/consensual.
- All rounds: No depictions of real-world harm, self-harm, or identifiable individuals; no hate speech or protected-class attacks.

## Consent & Disclaimers
- Every performer must consent to being roasted and to stylized, fictional harm within the AI-rendered world. Consent is recorded in submission metadata (`consent_roast`, `consent_virtual_harm`).
- Viewers are notified that depicted harm is fictional, stylized, and performed with consent; no real persons are injured.

## Fan Voting Rules
- Free base vote per verified user; paid credits allowed but capped (see monetization).
- Transparency: publish vote counts, fan_share, and bonus awarded after each battle.
- Anti-abuse: rate limits, device/IP checks, CAPTCHA on anomalies, audit log with reversible rollbacks.

## Leaderboard & Season Ops
- Source of truth: Season Points, then average Battle Score.
- Columns: Rank | Artist | Season Points | Avg Battle Score | Finisher Bonuses | Clean Runs | Fan Favor Bonus | Safety Pass Rate | W‑D‑L.
- Update after each judged battle; lock at playoffs start. Seeds taken from the locked table.

## Monetization
- Entry tiers: Free (view), Fan $5, Supporter $15, Patron $50 (perks, vote caps differ).
- Fan-vote credits: $0.25 each; packs 10/$2, 50/$9, 200/$30; non-transferable, expire end of season.
- Vote caps per event: Free 0, Fan 20, Supporter 100, Patron 200; max 5 votes/min; cooldown after 20.
- Sponsors: Tier A (Title, 1 slot), Tier B (Presenting, 3 slots), Tier C (Supporting, 10 slots). 30% of net vote revenue + 10% of net subs to prize pool.
- Audit log: user_id, event_id, sku_id, timestamp UTC, payment_id, amount, votes_granted, votes_cast, IP hash, device_id hash, operator_id for overrides; immutable 7-year retention.

## Safety & Escalation
- Pre-release sweep: prohibited content, minors, self-harm, extremism, PII/defamation, medical/legal claims.
- Rights check: licenses for all assets; talent/location releases; trademark clearance; takedown contact + SLA.
- Sora sensor checks: classifier scores, flags, watermark/traceability on; log prompt + outputs.
- Severity path: Sev1 immediate hold + exec escalation; Sev2 review <24h; Sev3 PM approval. Emergency stop + rollback; post-mortem required.