S8 / 04_FixedScale — COMPLETE NOMINAL HIPPARCOS FIELD CENSUS
Companion to manuscript v1.39 and S8 section 9.

Keep this directory alongside 03_Fields. The code imports that module's
archived catalogue, precession engine, selector and optimization helpers;
it does not download data or need a separately obtained paper.

From this directory, with Python 3.12 and requirements.txt installed:
  python audit_open_fields.py 1
  python audit_open_fields.py 0.5
  python verify_results.py

Each numerical run regenerates the complete candidate list before scoring.
The commands overwrite their own outputs in open_field_results/. To retain
released bytes, run in a copy. Runtime and environment fields in summaries
can differ across reproductions; score CSVs retain all numerical precision.
Allow several GB of memory; two half-year angular arrays alone use about
1.6 GB. No random sampling is used.

The released outputs include 19,939 unique quintets, 25,161 generating cap
intervals and both complete score grids. Every candidate uses one fixed
geometric role assignment; neither four-arm topology nor the horizon is an
entry requirement. The 283 four-arm and 256 cap-centred four-arm subsets are
reported separately. The latter exactly reproduces the original 03_Fields
membership, after canonicalizing its unordered residual pair.

RMS is measured in degrees against (7,18,4). Multiplying it by four gives
the old quarter-degree / sidereal-minute numerical representation. This
conversion does not remove the assumed scale. Free-scale E is the archived
relative dispersion, minimized separately for every candidate. It ranks
Cygnus 68th in the complete census and first in the four-arm subsets.

The fixed-scale horizon branch reoptimizes each candidate over every
admissible epoch segment, including boundary infima. Its denominator is
17,860, not the 13,910 whose unconstrained optimum itself passes the test.
The latter diagnostic is preserved in horizon_at_unconstrained_optimum.json.
The condition is necessary upper-culmination feasibility, not simultaneous
night-time visibility, a terrain model or evidence of actual observation.

Search window: astronomical years -8000 through +2000; BCE = 1 - year.
Grid refinement and endpoints use the original archived numerical helpers.
Agreement between two grids is numerical evidence, not an interval proof.
The verification tolerances report cross-grid precision and are not
archaeological uncertainties or thresholds fixed before seeing the data.

This is a retrospective nominal Hipparcos/J2000 extension. No four-panel
catalogue comparison or six-variant kinematic recensus is claimed for this
full fixed-scale ranking. Those earlier analyses concern the narrower
four-arm ratio family, as retained in 03_Fields and S8 sections 1–3.
Candidate counts and runner-up error ratios are not calibrated p-values.
