BAO Ruler Experiments

BAO Ruler Experiments This repository isolates one empirical question: Do late-time cosmological data prefer a fixed BAO ruler inherited from early-universe physics, or a free constant ruler calibrated by the late-time sky itself? The strongest current result is yes: the full late-time evidence stack prefers a free constant ruler over a fixed 147.09 Mpc anchor. Headline result From the packaged full late-time run: fixed-ruler baseline: logZ = -3769.272 free constant ruler: logZ = -3764.347 free-ruler advantage: ΔlogZ = +4.925 practical odds: about 140:1 preferred late-time ruler: 138.7 ± 1.9 Mpc A nearly identical free-ruler result is recovered in the CWCC-LT late-time branch, so the main empirical statement does not require accepting the full CWCC framework. What CWCC is, briefly CWCC stands for Causal-Window Cycle Cosmology. The motivating idea is that some cosmological tension may come from the way finite observations are translated into global distances, rather than from an immediate breakdown of local gravity. In plain language, the sky may behave normally nearby while the late-time inference rule used to build one rigid cosmic ruler is too strict. This repository does not try to package the entire CWCC programme. It packages the cleanest BAO-facing test that came out of it: compare a fixed inherited ruler against a free late-time ruler on the same data stack. What this repository is for This repository is about the BAO test itself. It packages: the exact scripts used to build the wiggle residuals and run the late-time model selection the hubble_systematics source package needed by those scripts the processed late-time data products used by the main BAO run the vendor BAO wiggle inputs used to construct the information-grade likelihoods the exact output packets for the headline run and the stepwise suite-build ladder the reconciliation documents showing why an earlier contradictory packet was traced to a fixed-ruler mapping bug a standalone manuscript focused on the BAO free-versus-fixed ruler result a reviewer reproduction seed that reruns the exact headline packet and checks the quoted numbers Main files Paper paper/bao_free_ruler_nature_style.tex paper/bao_free_ruler_nature_style.pdf Main run and audit docs docs/b0_bfixed_stepwise_suite_delta_logZ.md docs/b0_bfixed_reconciliation_referee_memo.md docs/b0_bfixed_reconciliation_suite_diff.md docs/b0_bfixed_unit_audit.md Exact headline packet outputs/late_time_model_selection_full_wiggles_packet_v2_20260303_173223UTC/ Stepwise ladder packets outputs/b0_bfixed_step_S0_wiggles_only_20260305/ outputs/b0_bfixed_step_S2_plus_bao_summary_20260305/ outputs/b0_bfixed_step_S3_plus_sn_20260305/ Wiggle extraction packet used by the headline run outputs/bao_entropy_event_wiggle_toy_realdata_packet_v11_allpk_plus_zhao_20260303_172044UTC/ Code archive Zenodo archive: https://doi.org/10.5281/zenodo.18908869 Reproducing the headline run Environment: python3 -m venv .venv source .venv/bin/activate pip install -e . Reviewer reproduction seed This is the easiest referee-grade path. It reruns the exact full-wiggle packet used for the headline BAO result and checks the recovered numbers against the locked reference payload. python3 scripts/reviewer_reproduce_bao_headline.py \ --out-dir outputs/reviewer_reproduction_seed_bao_headline The reviewer seed writes: reviewer_seed_summary.md reviewer_seed_results.json a fresh full-wiggle output packet inside the chosen output directory For a fast verification against the packaged reference payload without rerunning the full packet: python3 scripts/reviewer_reproduce_bao_headline.py \ --out-dir outputs/reviewer_reproduction_seed_bao_headline \ --skip-run Exact full-wiggle runner python3 scripts/evaluate_late_time_model_selection_full_wiggles.py \ --config configs/preregistered_model_selection/late_time_model_selection_full_wiggles_preregister_20260303_v2_all_info20.json \ --out-dir outputs/late_time_model_selection_full_wiggles_packet_v2_repro The packaged run in outputs/late_time_model_selection_full_wiggles_packet_v2_20260303_173223UTC/ is the exact reference result cited in the paper. How to read the result The most important interpretive points are: This is a late-time model-selection result. Wiggles alone are nearly neutral. The decisive sign flip appears only in the full late-time stack after adding calibrator poles. The result therefore supports freeing the late-time ruler calibration. Outside the broader CWCC motivation, the same empirical result can be read in standard language as evidence for a smaller sound horizon than the rigid 147.09 Mpc baseline. What this does not claim It does not prove CWCC microphysics. It does not prove that the early-universe acoustic picture is false. It does not replace local tests of GR. It does not say wiggle spectra alone already force a free ruler. The narrow claim is better and stronger: the late-time data package in this repository strongly prefers a free constant ruler over a fixed early-anchored BAO ruler.