USR-2026-0047 · Galactic dynamics

A Gaia wide-binary velocity analysis shows an apparent MOND-direction boost, but only under a clip-biased median estimator (residual 1.167). The unbiased mean gives residual 0.948 [0.939–0.958] — consistent with Newton. The boost is an estimator artifact of debiasing in the noise-dominated a₀ regime, where 28.7% of pairs are clipped.

KILLED residual 0.948 [0.939–0.958] — Newtonian (was 1.167 under clip-biased median) re-run & confirmed 2026-07-12 (R0 — founder CI; independence pending)

A wide-binary test that looked MOND-like under a clip-biased median collapses to Newtonian under the unbiased mean. Recorded as a self-kill: the boost was our own estimator, not gravity.

AUTOPSY REPORT — killed (self, the most honored outcome here)

Date: 2026-07-12 · Epitaph: "Killed by its own estimator audit: the MOND boost lived only in the clip-biased median. Under the unbiased mean the wide binaries are Newtonian (0.948). The 28.7% clip rate in the a₀ regime was doing the work."

Falsify-box — how to kill this claim

TRIGGERED MOND-direction boost survives switch from clip-biased median to unbiased mean estimator

Verification record — every quoted number, re-run 2026-07-12

quantity	measured	verdict
unbiased mean residual (headline)	0.948 [0.939–0.958] — Newtonian	MATCH
clip-biased median residual	1.167 — spuriously MOND-like	MATCH
RV-bound (3D-confirmed) residual	1.103 [1.062–1.148] — ~Newtonian/marginal	MATCH

Provenance

script: _wb_method3_definitive.py
script status: PRESENT
datasets: Gaia DR3 wide binaries · claim-scripts (downloadable from /data — run it yourself)
re-run: 2026-07-12

Edges

CONTRADICTS → Chae-type wide-binary claims of a MOND-direction gravitational anomaly: in this sample the boost is an artifact of the debiasing estimator in the noise-dominated a₀ regime, not a feature of gravity

Status, honestly

kill test fired: the MOND boost vanishes under the unbiased estimator. Status here is computed from evidence — the author cannot set it, and neither can we. Independent reproduction would move it; nothing else will.

§ Reproduction _wb_method3_definitive.py

The founder's own run of this script, captured verbatim. A match proves the result is reproducible; it is still R0 on this registry's independence rings — same code, so it cannot move a status. Only an outside run does that.

sha256 e995ab711d9f99d1a2728f58… size 13877 B exit 0 runtime 7.1s env python 3.12.2 · numpy 1.26.4 determinism byte-stable

A · Recorded founder's machine · 2026-08-02

loaded 214924 pairs, 214924 finite. gN/a0 in [0.013,192.0]
a0-regime (0.1-1): 35455 | high-g control (>10): 81813
rv_bound: 3515  (a0: 430, hi: 1768)

=== STEP 1: eccentricity calibration from vr_angle_deg (clean high-g control) ===
 beta  e_mean  ang_dist(chi2-like)
  0.5   0.600              0.00115
  1.0   0.667              0.00009
  1.5   0.714              0.00043
  2.0   0.750              0.00163
  3.0   0.800              0.00537
--> best-fit eccentricity index beta=1.0  (e_mean=0.667)
    NOTE: vr_angle is nearly uniform => weakly constraining; we also report thermal (beta=1) for robustness.

=== STEPS 2-5: forward model, debias, reference a0-regime to high-g control ===
  MAIN clean, thermal e (beta=1)     N_a0= 35455 N_hi= 81813  high-g(data/sim)=1.263  a0(data/sim)=1.472  RESIDUAL=1.165
  MAIN clean, calibrated e (beta=1.0) N_a0= 35455 N_hi= 81813  high-g(data/sim)=1.269  a0(data/sim)=1.480  RESIDUAL=1.167

=== STEP 6: bootstrap error on headline residual (Newton=1.0, MOND~1.15-1.4) ===
  thermal e                      residual = 1.165  [16-84%: 1.149-1.182]  => +9.89 sigma from Newton
  calibrated e (beta=1.0)        residual = 1.167  [16-84%: 1.151-1.185]  => +9.92 sigma from Newton

=== NULL test (fresh Newtonian sim as data; identical pipeline) — must be ~1.0 ===
  NULL residual = 0.996  (sanity: pipeline is unbiased if ~1.0)

=== STEP 7: rv_bound==True subsample (3D-confirmed bound, contamination-free) ===
  RV-bound, thermal e                N_a0=   430 N_hi=  1768  high-g(data/sim)=1.080  a0(data/sim)=1.196  RESIDUAL=1.107
  RV-bound, calibrated e (beta=1.0)  N_a0=   430 N_hi=  1768  high-g(data/sim)=1.074  a0(data/sim)=1.149  RESIDUAL=1.070
  RV-bound thermal               residual = 1.107  [16-84%: 1.045-1.190]  => +1.55 sigma
  RV-bound cal beta=1.0          residual = 1.070  [16-84%: 1.021-1.128]  => +1.37 sigma

  rv_bound clean: N_a0=430, N_hi=1768

=== Binned data/sim profile (calibrated e, clean sample) ===
     gN/a0       N  gam_data   gam_sim  data/sim
   0.1-0.3   11235     0.496     0.348     1.426
     0.3-1   24220     0.622     0.412     1.508
       1-3   36705     0.663     0.466     1.423
      3-10   58688     0.674     0.502     1.343
       >10   81813     0.664     0.523     1.269

=== HEADLINE: MEAN-based UNBIASED debias (the statistically correct estimator) ===
  MAIN clean residual = 0.948 [16-84%: 0.939-0.958] = -5.53 sigma  (N_a0=35455, N_hi=81813)
  high-g zero-point v_data/v_sim = 1.394 (model under-predicts; this is the divisor)
  RV-bound residual   = 1.103 [16-84%: 1.062-1.148] = +2.40 sigma  (N_a0=430, N_hi=1768)

=== VERDICT ===
  MEDIAN estimator (CLIP-BIASED, do not trust in a0): residual=1.167 -> spuriously MOND-like
  MEAN estimator (UNBIASED, headline): residual=0.948 [0.939-0.958] -> NEWTONIAN
  RV-bound (3D-confirmed, contamination-free): residual=1.103 [1.062-1.148] -> ~Newtonian/marginal
  -> CONCLUSION: data are consistent with Newton. The apparent MOND boost is an estimator artifact
     of debiasing in the noise-dominated a0 regime (28.7% of pairs clipped). MOND-direction boost NOT confirmed.

B · Yours not runnable in-browser

This script needs data files, a heavy dependency,
or more time than a browser tab should take.
Download it from /data and run it locally.

Share this claim X Bluesky LinkedIn Reddit HN Email