Matchmaking System Design

## ROLE
You are a matchmaking systems engineer who has designed skill-based matchmaking for competitive multiplayer games. You understand rating algorithms (Elo, Glicko, TrueSkill), queue theory, and the complex balancing act between match quality, wait times, and player experience.

## OBJECTIVE
Design a complete matchmaking system that creates fair, balanced matches quickly while accounting for skill level, latency, party sizes, and player preferences.

## TASK
Create a comprehensive matchmaking system:

### Skill Rating System

**1. Rating Algorithm Selection**

**Elo Rating:**
- Simple, well-understood, works for 1v1
- K-factor tuning for volatility control
- Limitation: Designed for 1v1, not teams
- Best for: Chess-like games, 1v1 fighters

**Glicko-2:**
- Adds rating deviation (confidence interval)
- Accounts for time since last game (rating decay)
- Better for infrequent players
- Best for: Games with varying play frequency

**TrueSkill / TrueSkill 2:**
- Designed for team games and free-for-all
- Separates individual skill from team performance
- Handles partial play and different team sizes
- Best for: Team-based competitive games

**Custom Composite Rating:**
- Combine multiple factors beyond win/loss
- Performance metrics (KDA, objective score, damage dealt)
- Role-specific ratings
- Consistency weighting
- Best for: Complex team games with roles

**2. Rating Calculation**
- Initial rating assignment for new players
- Placement match system (10-25 games for calibration)
- Rating adjustment formula per match
- Win/loss, margin of victory, individual performance weighting
- Party MMR adjustment (groups rated differently than solo)
- Smurf detection and accelerated placement
- Rating decay for inactive players
- Season reset strategy (hard, soft, proportional)

### Queue Management

**1. Queue Architecture**
```
Player joins queue
  -> Validate eligibility (rank, level, behavior score)
  -> Add to pool with preferences (region, mode, role)
  -> Matching algorithm runs periodically
  -> Propose match to players
  -> Accept/decline handling
  -> Match confirmed -> Launch game server
```

**2. Matching Algorithm**
- Priority-based matching:
  - Priority 1: Skill proximity (tight at first, widens over time)
  - Priority 2: Latency (geographic proximity)
  - Priority 3: Wait time fairness
  - Priority 4: Party size balance
  - Priority 5: Role composition (if applicable)
- Search range expansion over time
  - 0-30 sec: Tight skill range (plus or minus 100 MMR)
  - 30-60 sec: Medium range (plus or minus 200 MMR)
  - 60-120 sec: Wide range (plus or minus 400 MMR)
  - 120+ sec: Very wide with match quality warning
- Maximum wait time before fallback match

**3. Team Balancing**
- Equal average MMR between teams
- Minimize MMR spread within teams
- Party together on same team
- Role distribution balance
- Experience level distribution
- Avoid repeated matchups (recent opponents filter)

### Regional & Infrastructure

**Server Selection:**
- Geographic server regions and coverage
- Player latency measurement and server ranking
- Cross-region matching rules and thresholds
- Peak time handling per region
- Low-population region merging strategy

**Infrastructure:**
- Matchmaking server architecture
- Game server allocation and scaling
- Session management and cleanup
- Database design for player profiles and match history
- Redis/cache layer for real-time queue data

### Anti-Abuse & Fair Play

**Smurf Prevention:**
- New account detection heuristics
- Accelerated MMR adjustment for new accounts
- Performance anomaly detection
- Phone/identity verification requirements
- Hardware/IP based detection

**Queue Abuse Prevention:**
- Dodge penalty system
- AFK detection and penalties
- Queue sniping prevention (streamer mode)
- Win trading detection
- Boosting detection (party with much lower-rated players)

**Behavior Score Integration:**
- Toxic players matched with each other
- Good behavior rewarded with better match quality
- Report system integration
- Abandonment penalty system

### Player Experience

**Visible vs. Hidden MMR:**
- Hidden MMR for matchmaking accuracy
- Visible rank for progression feeling
- Rank tier system (Bronze through Champion)
- Promotion and demotion mechanics
- Season rewards and reset

**Match Quality Indicators:**
- Pre-match predicted balance score
- Post-match feedback collection
- Match quality metrics tracking
- Player satisfaction correlation with match parameters

### Analytics & Tuning
- Match quality distribution tracking
- Wait time percentile monitoring
- Balance accuracy analysis (predicted vs. actual outcomes)
- Queue abandonment rate
- Rating distribution curves
- A/B testing framework for algorithm changes