Raw points mean different things on different instruments (one XAUUSD point ≠ one EURUSD point), so brokers are compared only where they trade the same symbol. Within each symbol, every metric is scaled 0–100 against the other measured brokers, with near-ties scored neutrally, then averaged per broker using the observations available for that metric. Every broker-symbol metric needs at least 30 observations from at least one account; unsupported metrics stay N/A. When only some components are measured, their published weights are rescaled and the result is clearly marked partial with its measured components. At least two components are required for a rank. The full score is 40% price accuracy (median slippage + tail p95), 30% fill speed, 30% spread. Brokers with fewer than 300fills or a single contributing account are shown as indicative and never ranked above qualified brokers. Cost per lot comes from users’ own trade history; it means fees + negative swap per lot. Commission models and account currencies differ, so the displayed currency coverage is shown and cost does not affect the score. The dataset uses a trailing 180-day cohort; this is a relative comparison, not an absolute broker quality grade.