WHAT THIS IS
Six crews of tiny minds work the same coalfield. Each has ONE setting:
how much roof safety it insists on before it will cut. Everything they
earn or lose follows from that number.
The ground is real. 520 surveyed underground workings from the
West Virginia Geological Survey, mined 1907–2008, across 15 coal
seams stacked over each other. The pale shapes are what previous crews
already took. The dark ground between them is what is left.
The physics is real. Pillar loading by tributary area, pillar
strength by the Obert–Duvall relation, safety factor as strength
over stress. Deeper ground needs fatter pillars, and fatter pillars mean
less coal. There is no free lunch.
Nobody knows where the coal is. What is mapped is what earlier
crews already took out. Beyond that edge is unknown — which is true
in reality, and is the whole hunt.
WHAT A CREW CANNOT DO
A crew cannot read the true strength of the rock. It reads the roof the
way a real miner does — sounding it, watching how the coal cuts
— and that reading is noisy. It decides on what it can see,
and the ground settles up afterwards.
Roof bolts help, cost money, and hold about four times in five.
Support reduces the chance of a fall. It does not abolish it.
WHAT THE SIM FOUND
Run it and the middle wins — on money and on safety at the same
time. Measured across all six seams from 60 m to 285 m of cover,
the best crew sits at SF 1.4–1.6, never at either extreme.
Below about SF 1.4 there is no tradeoff at all. The reckless crew
does not buy coal with the safety it gives up. It backs out of workable
ground it misread, drops roofs on the ground it does take, and ends with
less coal AND less money AND the collapses. Giving up safety past that
point loses on both axes.
And the two failures are not symmetric. Beating the reckless crew
is worth +158% to +384% depending on the seam. Beating the timid one is
worth +5% to +12%. Excess caution is cheap. Recklessness is ruinous.
The ranking is stable, not a short-run artefact: SF 1.6 leads at 200
panels and still leads at 4,000, with the gap holding near 2.7×.
The reckless crew never catches up.
MODELLED, NOT MEASURED
The workings, the seam, the year and the mine name come from the survey.
Depth and seam thickness are modelled — those layers carry
plan geometry and no elevation, so the spacing is a stated constant
rather than a measurement. Said plainly so nothing here is mistaken for
data it is not.