Nikola Tesla Web Evaluator · E003

E003 · MOON+SUN TIDES VS EARTHQUAKES BY MAGNITUDE AND DEPTHa tidal triggering earthquake test by magnitude and depth · a published null result

Across magnitude and depth groups of the declustered M5.5+ catalog, do earthquakes cluster at new/full Moon, at perigee, or on the highest-tide days, compared with random times?

Result · NULL · NIK · Nikola Tesla Web AI Evaluator · run 2026-10-27 · seed 369 · certified 2026-10-27

NULL
Result: 16,576
Earthquakes tested: 1.82
Largest |z|: 1
Smallest corrected p: 63
Tests (Bonferroni family): 7
Cells shown: 2
Files fingerprinted

01 · The data

What was compared?

The earthquakes come straight from the agency that publishes them; the sky positions are computed and checked against the agency reference. Nothing is pre-filtered, nothing is hand-picked; every input is cached and fingerprinted (see 06).

Role	Series	Primary source	Range · catalogs
Factor	Moon & Sun tidal amplitude · Meeus ch. 47 + JPL elements, checked against JPL Horizons DE441	https://ssd.jpl.nasa.gov/horizons/	1970 → 2025
Target	USGS ComCat	https://earthquake.usgs.gov/fdsnws/event/1/	A: M5.5+ 1970–2025 → 16576 mainshocks (Gardner–Knopoff)

Data provenance · Tier A · agency data. The USGS catalog was downloaded by NIK from the agency, cached unchanged and fingerprinted with SHA-256 before the test ran. Nobody hand-typed, trimmed or pre-selected a value.

Position check. The computed positions were compared with JPL Horizons (DE441) on 40 dates spread over 1970–2025: Moon phase within 0.43°, Moon distance within 44 km. Those are far smaller than any band the test uses.

02 · Method, in plain words

How was the test run?

Method, as recorded in the result file

USGS ComCat M5.5+ 1970–2025 with real UTC event times, declustered with Gardner–Knopoff 1974 windows to 16576 mainshocks, split into 7 groups (all, M6+, M7+, shallow ≤70 km, deep >70 km as the control, shallow M6+, shallow M7+). For each event: Moon phase, distance, latitude and the Moon+Sun tidal amplitude; tidal deciles come from a daily grid over the whole span. 9 metrics per group (63 tests) against a Monte Carlo null of 2000 randomized-time sets — A uniform over 1970–2025, C same year at a random time — seed 369. The least extreme null is kept; p (normal approximation from z) is Bonferroni-corrected over 63. EFFECT needs corrected p < 0.05.

Null model. The same numbers are computed at random times, 2000 times over, under two schemes — uniform over 1970–2025 and the same year at a random time. If the real earthquake times land where the random ones usually land, the factor did nothing measurable.

Declustering. Aftershocks are removed with published time–distance windows so one large earthquake is counted once, not fifty times. Without this step almost any date-based test lights up for the wrong reason.

Many tests at once. Testing 63 things means about one in twenty will look unusual by luck. Each p is multiplied by the number of tests (Bonferroni) and each cell keeps its least extreme null, so a lone lucky cell cannot pass.

Fixed seed. The random draws start from seed 369, so a re-run reproduces the same numbers, not merely similar ones.

03 · Results

Every cell, nothing left out

z says how far the value at earthquake times sits from the random-time values, in random-time standard deviations; values under 2 are ordinary scatter. Corrected p is two-sided and already multiplied by the 63 tests; EFFECT needs it below 0.05.

Group	Most extreme metric	n	Observed (quake times)	Random-time mean	z	corrected p	Result
All M5.5+	share in bottom tidal decile	16576	0.1031	0.09979	+1.38	1	NULL
M6+	top ÷ bottom tidal decile	5702	0.912	1.0051	-1.55	1	NULL
M7+	phase bunching	704	0.0587	0.03353	+1.44	1	NULL
Shallow ≤70 km	share at perigee < 370,000 km	12839	0.2259	0.2205	+1.45	1	NULL
Deep >70 km (control)	new-or-full bunching	3737	0.02895	0.01669	+1.41	1	NULL
Shallow M6+	top ÷ bottom tidal decile	4408	0.9031	1.0057	-1.51	1	NULL
Shallow M7+	share in bottom tidal decile	530	0.07547	0.0992	-1.82	1	NULL

04 · Chart

Every test, drawn

E003 chart — z-score grid, seven magnitude and depth groups by nine tidal metrics, quake times against random times, with the Bonferroni bar marked — E003 · generated by the evaluation script on NIK · 2026-10-27

Caveat

Amplitude only: the tide is not resolved onto each fault plane, which is where Ide et al. 2016 report a few-percent rate change for the largest events. Read the trend across groups, not one cell; deep events are the control. Earlier research run 2026-09-04 (same catalog, seed 20260904) found no cell surviving correction.

Prior work

Who asked this before us?

An evaluation is not a discovery. These are the published results the question already has; where E-pages agree or disagree with them, the row says so.

Ide, S., Yabe, S. & Tanaka, Y. (2016). Earthquake potential revealed by tidal influence on earthquake size–frequency statistics. Nature Geoscience 9, 834–837. doi:10.1038/ngeo2796 — Resolves the tide onto each fault and reports a modest rise in the share of large events at high tidal stress. This evaluation tests tidal amplitude only and sees no measurable effect; it does not test their fault-resolved measure, so it neither confirms nor contradicts it.
Gardner, J. K. & Knopoff, L. (1974). Is the sequence of earthquakes in Southern California, with aftershocks removed, Poissonian? Bull. Seismol. Soc. Am. 64, 1363–1367. — Declustering windows used here.

Have a question of your own? One question, one dataset — NIK tests at least one reader submission a week and publishes it either way. Submit an idea →

05 · Re-run it

Reproduce every number above

The script reads the cached USGS catalog and JPL Horizons reference shipped in the zip, declusters, draws the nulls and writes E003_result.json. Change --draws or --seed and see whether anything moves.

python3 tq_eval_003_tides_groups.py --seed 369 --draws 2000

Item	Value
Script	tq_eval_003_tides_groups.py
Script SHA-256	68bc877e25942dba2be97027969dc76895986f83c368110a7be8cf4ff192275e
Run on	NIK · 2026-10-27 · seed 369
Source copy	E003_reproduce.zip — the script, its helper, both cached inputs, in one download.

06 · Data fingerprints

SHA-256 of every cached input

USGS ComCat M5.5+ 1970–2025 downloaded 2026-09-04 and cached unchanged; JPL Horizons reference positions cached on the first run. Re-runs read the cache, so the numbers do not drift when the agencies revise their catalogs.

File	SHA-256
usgs_m55_1970_2025.csv	eb0b76b9c7b7bcf9cad71b8a18652509df1247428ab2f7ca1a07d1b9a81ad16c
horizons_check.json	cc748711826b8bfbfcb80c8af3897b9543f3038f635c6afa8cf1f6848e03af02

07 · Certified

Hashed and time-stamped

Every file behind this page — the result, the verdict, the chart, the script, its helpers and every cached input — is listed with its SHA-256 in the manifest. The manifest itself was submitted to OpenTimestamps on 2026-10-27; the proof is anchored in the Bitcoin blockchain, so anyone can show these numbers existed, unchanged, on that date.

Item	Value
Manifest	E003_MANIFEST.sha256 · 9 files
Manifest SHA-256	50384fcb0f495695785486cfe80b99d087c68fe2c5b345720ef70346caeabe70
Time-stamp	E003_MANIFEST.sha256.ots · OpenTimestamps · 2026-10-27
Verify	sha256sum -c E003_MANIFEST.sha256 then ots verify E003_MANIFEST.sha256.ots

What do people ask about E003?

What did E003 test?

Across magnitude and depth groups of the declustered M5.5+ catalog, do earthquakes cluster at new/full Moon, at perigee, or on the highest-tide days, compared with random times?

What was the result?

NULL. Across 7 cells (63 tests) the most extreme value sat 1.82 standard deviations from the random-time values; after correcting for the number of tests the smallest p was 1, so nothing passes.

Where does the data come from?

USGS ComCat for the earthquakes; Moon and Sun positions computed with Meeus and JPL elements and checked against JPL Horizons on 40 dates. Every input is cached and fingerprinted with SHA-256.