London regional baseline experiment, 10 October 2026

Read the HTML report and both protocol JSON files first. These are exploratory, previously inspected outcomes. No website deployment is part of this archive.

Dependencies: Python 3, numpy, scipy, matplotlib. Original versions are in frozen-inputs.json.

Reproduce in an extracted copy:
OPENBLAS_NUM_THREADS=1 python3 expanded.py verify
OPENBLAS_NUM_THREADS=1 python3 expanded.py point
OPENBLAS_NUM_THREADS=1 python3 expanded.py holdout
OPENBLAS_NUM_THREADS=1 python3 expanded.py bootstrap
OPENBLAS_NUM_THREADS=1 python3 expanded.py summarise
OPENBLAS_NUM_THREADS=1 python3 present.py

Do not rerun freeze: protocols and hashes are already archived. The original V1 point results are preserved as an input to the recorded amendment. The compressed bootstrap-refits archive contains all 399 draw records; checkpoint directories are excluded from this ZIP, so rerunning bootstrap recomputes the draws. Failed-draw-sensitivity.json is deterministic post-processing of the archived draws; see its method field and report.

Input source lineage and coverage are retained in source/frozen-design.json. Original source study files are copied, never modified. Predictions are conditional on realised target-year mileage.
