benchmarks

Generate FlawChess population-level benchmarks from the prod or local dev database — per-user score-gap (endgame vs non-endgame), endgame Conversion/Parity/Recovery rates and composite Endgame Skill distribution bucketed by ELO (500-wide) and time control, Endgame ELO vs Actual ELO gap distribution per (platform, time-control) combo, time-pressure stats at endgame entry (avg clock diff, net timeout rate), and time-pressure-vs-performance curves across time controls. Use this skill whenever the user asks about endgame benchmarks, neutral zones, gauge ranges, "what's typical", baseline distributions, calibrating thresholds, comparing time controls, deciding whether to collapse time controls, setting conversion/recovery/parity/skill ranges, or calibrating Endgame ELO timeline expectations. Trigger on phrases like "benchmark", "benchmarks", "baseline", "neutral zone", "gauge range", "distribution of rates", "how are rates distributed", "score gap distribution", "conversion distribution", "endgame skill distributi