| Moon | Sec | Tier | Composition | Total |
|---|
How the estimate works
Data. Every killmail since 7 June 2024 is read from the zKillboard R2Z2 daily raw archives and Metenox Moon Drill losses are kept. A drill is matched to the nearest moon in its system by position (drills sit 1,000–10,000 km from the moon centre).
Drill mechanics. Each hour a Metenox extracts 30,000 m³ of moon ore split by the moon's ore fractions, and reprocesses it at 40% (static data metenoxMoonDrill). For a moon ore (10 m³, 100-unit batches) that is 12 × f × q units of each material per hour, where q is the SDE reprocessing quantity. The hourly yield of each ore/material is floored to a whole unit.
Composition. Moon materials in the bay (dropped + destroyed) are fitted with non-negative least squares. Each ore yields one material that no lower tier yields, so the system is triangular and the fit is exact.
Absolute fractions. Because hourly yields are whole numbers, every material count is a multiple of the hours since the bay was last emptied, H. Candidate values of H are divisors of the counts' GCD, checked for an exact floored solution. Moon totals in EVE are whole tenths (0.5–1.0), and a Bayesian step picks the most likely H, pooling every kill of the same moon. That gives fractions to about ±0.2%. If the owner hauled out only some materials, the bay holds two accumulation periods. The solver then splits the materials at a tier boundary, and accepts the result only when exactly one split fits. When nothing fits, only relative shares are shown.
Validation. For 2,145 pairs of kills of the same moon, taken days to months apart, the recovered hourly yields agree exactly. Recovered totals sit within 0.003 of a whole tenth for 99% of moons.
Limits. Higher-yield ore variants (Brimful, Glistening, …) have the same ratios as their base ore, so they show up as extra base ore. R4 / R8-only moons can be ambiguous between a total of t and t/2; the confidence column says when.