Coal Method Comparison -- Threshold-Flag vs. Single Centroid vs. 4-Rank
Head-to-head comparison of the three approaches discussed for adding Coal to the empirical classifier: (1) Crain's threshold-flag rule alone, (2) Coal as ONE mineral-style endpoint (Bituminous, same treatment as Sandstone/Limestone/Halite/Anhydrite), and (3) the full 4-rank mixing model (the 'Coal Identification & Rank Classifier' card). Run on the FULL depth range of every well with a SANDCOUNT Coal pick, not just inside the picks -- gives real precision, not recall alone.
What "accuracy" means here: Same caveat as the other Coal card: SANDCOUNT's 'Coal' picks are not verified geological ground truth (undocumented Lloyd's Register classification). Precision/recall here measure agreement with that provisional label, on the FULL depth range of coal-bearing wells (not just inside the picks) -- a method that fires everywhere would score badly on precision even if it never missed a real SANDCOUNT pick, which is the point of measuring both, not recall alone.
Last run: 2026-08-02 08:49:10 UTC
How this result was generated
SANDCOUNT 'Coal' pick"] full["Full depth range
(inside AND outside picks)"] m1["Method 1: Threshold-flag
(3-of-5 Crain rule)"] m2["Method 2: Single centroid
(Bituminous endpoint only)"] m3["Method 3: 4-rank mixing model
(nearest of 4 endpoints)"] ext["External coherence:
precision/recall/F1 vs SANDCOUNT"] int_["Internal coherence:
tightness of each method's
OWN positive calls"] wells --> full full --> m1 --> ext full --> m2 --> ext full --> m3 --> ext m1 --> int_ m2 --> int_ m3 --> int_
Hover a step for a plain-language explanation.