Methodology: Runners ranked by weighted performance index (WPI) from comparable past starts. WPI = benchmark_average ÷ pace-adjusted estimated time. Top 3 per race. WPI > 1.00 = faster than par.   Time estimation: Winners use recorded time directly. Non-winners reverse-engineered: winner_time + (margin × spl), where spl = (winner_time ÷ distance) × 3.2 — self-adjusts for track conditions via race pace.   Pace adjustment: Leader who won −0.4% (set own pace). Strong closer from rear +0.7% (overcame traffic & ground). All others: no adjustment.   Condition matching: ±1 step = 85% weight — primary ranking (G4↔S5 acceptable). ±2 steps = 60% weight — supplementary context only (G3↔S5, use with caution). >2 steps = excluded.   Distance matching: Graduated weight: exact=100%, ±200m=90%, ±400m=70%, ±600m=50%, beyond=30%. Never excluded — absence at a distance is neutral, not negative.   Confidence: HIGH = 3+ primary runs + proven at distance. MEDIUM = 2+ primary runs or adjacent distance evidence. LOW = 1 primary run or supplementary only. INSUFFICIENT = no comparable data.