Player-of-One Personalized Dynamic Difficulty

You are a lead game data scientist designing a personalized, player-of-one dynamic difficulty system: Game type: [INSERT - puzzle RPG, shooter, idle, strategy PvE] Platform: [INSERT - mobile, PC, console] Adjustment levers: [INSERT - enemy stats, level selection, economy, assists] Player population: [INSERT DAU] Cover the full pipeline: 1. Design - target experience pillars, where adaptation is invisible vs optional, difficulty philosophy, anti-frustration and anti-boredom principles, fairness commitments. 2. Player modeling - per-player skill estimation, mastery curves, playstyle and preference inference, frustration/engagement/boredom classifiers, confidence and cold-start handling. 3. Technology - real-time feature pipeline, online inference latency, decision service architecture, rule guardrails around ML outputs, config-driven levers. 4. Adjustment engine - multi-armed or contextual bandit over difficulty configurations, reward and pacing adaptation, assist features, cooldowns to avoid oscillation. 5. Data and measurement - event schema, counterfactual and holdout design, D0/D7 retention and spend outcomes, flow-state proxy metrics, segment drilldowns. 6. Testing - simulator validation, shadow mode, canary rollout, exploit tests (gaming the system, boosting), accessibility and subjective-difficulty surveys. 7. Launch - player-visible controls, transparency messaging, whitelists for competitive modes, staged rollout by country.