Transits as Event Triggers: A Three-Stage Test of Classical Gochara Timing — Raw Transits, the Aṣṭakavarga Refinement, and the Moon-Degree Layer — on Certificate-Grade Lives

The classical answer to "when, within the era, does the event actually arrive" is gochara — the doctrine that Saturn, Jupiter, Rāhu and Ketu, transiting sign-wise from the natal Moon (and, in modern practice, degree-wise against natal points), supply the trigger that the daśā period promises. The daśā layer was the subject of our first rectification backtest and it failed: under two formalisations it recovered no birth-time signal from day-granular events, and where it deviated from chance it deviated in the wrong direction. That result left exactly one classical layer untested — the trigger layer itself. This study tests it in three stages, each pre-registered before its run: raw gochara against the same 23 certificate-grade lives, the classical Aṣṭakavarga refinement of it, and finally the only transit channel that planetary motion physically allows to resolve an hour — the Moon — against a fresh registry of nineteen independently minute-dated events on eight AA-anchored charts. The first stage produced a candidate signal that did not survive its own pooled nulls — a candidate, not a confirmation. The second showed the classical refinement actively degrades ranking. The third found zero of nineteen events on a Moon–significator conjunction at the documented minute (p = 1.0 against the hour-shuffle null). No tested layer shows a confirmed positive timing signal — the Aṣṭakavarga and Moon layers are falsified; raw gochara remains inconclusive at this power. The one rung left untested, and the resolution ladder that shows why, are set out at the end.

Abstract

The companion study (Birth-Time Rectification: A Blinded 23-Subject Backtest) falsified the daśā–lordship core of event-based rectification on day-granular events, under both an additive and a likelihood engine. Classical practice, however, never uses daśā alone: it uses gochara — transits — as the trigger that fires within the era, judged sign-wise from the natal Moon (Janma Rāśi) with Saturn, Jupiter, Rāhu, Ketu and secondarily Mars as the significant long-duration bodies, and refined by Aṣṭakavarga or by exact-degree conjunctions. This study is the pre-registered falsification of that trigger layer, run in three stages on the same certificate-grade (AA) cohort and machinery.

Stage 1 (raw gochara, day-granular): the classical channels — transit significators conjunct (±2°) the candidate lagna, the natal Moon, and the event's natal significators, plus the sign-from-Moon rule — were scored rarity-weighted over 1,440 candidate minutes per subject for the 23-subject cohort's primary event set (marriage, childbirth, accident/death; 3–8 events each). Result: 0/23 recoveries within ±30 minutes (closest Diana Δ60, Stéphanie Δ64); the recorded time's cohort-mean percentile is 17.2 — strikingly better than the daśā likelihood's 69.5, with 4/23 subjects in the top 5% of their own day (binomial p = 0.026) — but its own age-window null also ranks high (mean 15.7), and the pooled real-vs-null comparison is not significant (p = 0.73 vs the age-window null; per-batch p = 0.40–0.93). Per the pre-registered gate the verdict is candidate, not confirmed. The transit layer is frame-robust where daśā was frame-fragile: the Lahiri frame cell returns byte-identical statistics for all four top-5 subjects, a structural consequence of testing angular separations, which are invariant under a uniform ayanāṁśa shift.

Stage 2 (the Aṣṭakavarga refinement): the classical filter — BAV ≥ 5 / SAV ≥ 30 sign thresholds from the BPHS contribution table (frozen from the Jagannātha Hora implementation; row sums and the 337 SAV total verified) — was added to the engine as a third arm and tested against the v1 reproduction control. The AV layer is a ~2-hour step function by construction (BAV/SAV depend only on sign positions), so it cannot resolve minutes; measured on the AV-only arm it yields 8–10 distinct scores all day with 119–156-minute flat maxima. Adding it degrades the v1 ranking (cohort mean percentile 45.5 vs 17.2; top-5% 1/23 vs 4/23), and four of the five nominal AV-only top-5% "hits" are confirmed plateau artefacts (the recorded minute sits inside a wide flat maximum — pct 0.0 by tie). The gate is not met on any arm (pooled p = 0.65–0.73).

Stage 3 (the hour-level layer, the Moon): slow planets physically cannot resolve hours — Saturn at 0.25° orb makes a ~15-day window and Jupiter ~6 days — so the only hour/minute-resolving transit channel is the Moon (13.2°/day; a 0.25° orb window lasts ≈ 54 minutes). We built a fresh evidence class for it — nineteen events with independently documented minutes (death-certificate hours, palace birth bulletins, marriage service clocks) across eight subjects whose birth anchors are all AA, every event time AA/A with two independent sources, every timezone/DST offset audited — and asked, on charts whose birth time is already known, whether each documented event minute falls on a Moon conjunction (≤ 0.25°) with a classically grounded natal significator target. Result: 0/19 fires against an hour-shuffle null that expects 0.07 — p = 1.0 by 20,000-replicate permutation and by exact Poisson-binomial; identical at 0.5° and 0.1° sensitivity orbs. (When the observed count is zero — the minimum possible — a one-sided p-value testing "real ≥ null" is guaranteed to be 1.0 by construction; the informative companion fact is that a zero-fire result occurs ≈ 93% of chance draws when the null mean is 0.07, i.e. it is expected, not anomalous.) The sign-level Moon channel (Moon in the 8th from natal Moon or lagna, M-8) fires 3/19 against heterogeneous per-event sign-occupancy rates (mean 0.185; many events have q = 0, two have q = 1.0); the exact Poisson-binomial p = 0.90 reflects this heterogeneity, not a uniform Binomial — also null. Stage 2 of the timed-event spec (minute-level birth-time recovery) is not entered.

Verdict of the programme so far: no tested layer shows a confirmed positive signal: the daśā era is falsified under two formalisations, raw gochara is a candidate that fails its pooled nulls (inconclusive at this power), the Aṣṭakavarga refinement is null and degrades ranking, and the Moon-degree layer is null at every orb. The one remaining rung, event-moment angles (~1°/4 min), is untested because it requires minute-dated events at a known birth site, a stronger evidence class than the public record has so far yielded for any certificate-grade cohort. What this does and does not prove, in full, is stated at the end.

Why transits, and why this is the natural next falsification

Bṛhat Parāśara Horā Śāstra and the practice tradition teach that daśā describes the era of a result and gochara supplies the trigger within that era: marriage arrives when Jupiter or Venus transits significator points; death and serious accident when Saturn, Rāhu or Mars transit the lagna, the Moon or the māraka significators; the canonical Sade Sati (Saturn over the 12th/1st/2nd from the Moon) is the archetype of a transit-marked period. Classically, gochara is judged sign-wise from the natal Moon with Saturn, Jupiter, Rāhu, Ketu and (secondarily) Mars as the significant bodies, refined by Aṣṭakavarga; modern software additionally uses degree conjunctions (a slow planet at the same longitude as a natal point). Both forms are testable with the same cohort and machinery as the daśā study — and the trigger layer is precisely the one that daśā–lordship scoring never touched. A failure of daśā alone therefore says nothing about the fuller art; a failure of daśā and gochara, raw and refined, says a great deal more. This is the logic of the three stages below, each specified and committed to the study repository before any run.

Subjects, evidence and protocol (frozen from the daśā study)

The same 23-subject cohort and payloads as the rectification backtest (certificate-grade AA anchors; Churchill re-rated A and flagged; Einstein's AA verified; Elizabeth II flagged as a weak anchor). The primary event set is frozen to the classically unambiguous marriage, childbirth and accident/death events (3–8 per subject, mean ≈ 4.5); career, move and spiritual events are excluded from the primary analysis because their transit signification is textually vaguer. Pushya Pakṣa ayanāṁśa (the NativeAstro production frame), whole-sign houses, mean nodes, Swiss Ephemeris, geocentric, one-minute candidate grid, no LLM anywhere in a scoring path, and the recorded birth time never supplied to the engine. Decision statistics are identical to the likelihood study: the recorded minute's percentile among the 1,440 candidates, with winner-within-±30-minutes as a pass rule for reportability only; the pre-registered coverage gate requires ≥ 4/23 subjects in the top 5% of their own posterior and a cohort mean that beats the age-window null at p < 0.01 and survival of the Lahiri frame cell. Nulls: age-window (primary, life-course-preserving), date-only, and joint type+date shuffles, 60 draws per null kind per subject.

Stage 1 — raw gochara on the day-granular cohort

Four channels, all binary and rarity-weighted so that all-day "hits" self-deflate: T-lagna (a transit significator of the event type conjunct ±2° the candidate lagna degree on the event date), T-moon (conjunct ±2° the natal Moon), T-sig (conjunct ±2° the natal degree of the event's classical significator planets), and S-moon (the classical sign rule from the natal Moon: for marriage, Jupiter/Venus in the 7th; for childbirth, Jupiter in the 5th; for accident/death, Saturn/Mars/Rāhu in the Sade Sati and Dhaiya positions). Significator sets fixed in the spec: marriage {Jupiter, Venus}, childbirth {Jupiter}, accident/death {Saturn, Mars, Rāhu}; orbs ±2° pre-registered with a 1°/3° sensitivity cell.

Reading. The recorded-minute percentiles are qualitatively unlike the daśā layer: the transit model ranks real recorded times far higher (cohort mean 17.2 vs 69.5) and ten subjects sit in the top ~10% of their own day. That is the direction a real trigger signal should point, and it justified the refinement stage rather than closure. But the pooled comparison against a life-course-preserving null is not significant (p = 0.73), zero subjects recover within ±30 minutes, and the four top-5% subjects include the same names whose daśā scores were chance-shaped — so per the pre-registered rule the state is candidate, not confirmed.

Frame cell. The four top-5 subjects were re-run under Lahiri after fixing a frame-consistency bug (transits are now computed inside the patched block). Pushya 22.18° vs Lahiri 23.32° at the Obama birth JD — the patch is real. All four survive with byte-identical statistics (percentiles and top-5 flags unchanged). This is structural, not lucky: gochara channels measure angular separations and sign positions from the natal Moon, both invariant under a uniform ayanāṁśa shift — the mirror image of the daśā lordship layer, where a 1.1° shift moved lagnas over whole-sign boundaries and changed the winner on 23/23 subjects.

Stage 2 — the Aṣṭakavarga refinement

Aṣṭakavarga is the classical filter on gochara: each planet's transits are weighted by a contribution table (bindu counts) so that only transits through fortified signs are read as triggers. We operationalised the JHora-default BPHS encoding — the contribution table frozen from the Jagannātha Hora implementation (PyJHora), with per-receiver row sums verified against the classical invariants (Sun 48, Moon 49, Mars 39, Mercury 54, Jupiter 56, Venus 52, Saturn 39; SAV = 337) and all six school-variant toggles confirmed no-ops at JHora defaults — and added it to the engine as a third arm alongside the v1 reproduction control, with the sign-threshold rule BAV ≥ 5 / SAV ≥ 30, all three arms run over the same cohort and nulls.

The v1 control reproduces the published v1 cohort exactly (mean 17.2, top-5 Obama/Ali/ Turner/Jordan) — the shared machinery is validated end to end. The AV finding is then unambiguous: adding the classical refinement materially worsens minute ranking. Mechanically this is what the resolution constraint predicted — BAV/SAV depend only on sign positions, so over a candidate day the AV grid changes only when the lagna (or rarely the Moon) crosses a sign boundary. Measured on the AV-only arm: William shows 8 distinct score values all day with a 119-minute flat maximum; Stéphanie 10 distinct values with a 156-minute flat maximum. Four of the five nominal AV-only top-5% rows (William, Stéphanie, JFK, Churchill) have pct 0.0 — the recorded minute sits inside a ~2-hour flat maximum, which the top-5% rule then calls a "hit". The spec's evidence rule 4 anticipated exactly this plateau-alignment artefact, and it does not survive investigation: a 119–156-minute tie plateau contains the recorded minute about as often as any other minute. Trump's AV row (percentile 1.3, Δ93) is a genuine low percentile rather than a tie, but it is one subject against a pooled p = 0.65. The classical claim that Aṣṭakavarga is the necessary filter on gochara is therefore not supported at the resolution this event set can test — AV may still have value for sign- or season-level judgement, but it cannot contribute to minute-level recovery, and bolted onto a degree-based engine it is actively harmful.

Stage 3 — the hour-level layer: the Moon at documented event minutes

Both earlier stages shared the same limitation: their events were dated to the day, and the classical bodies they track are slow. The physics fixes what the slow-body layer can ever claim — Saturn and Jupiter move ≈ 0.033°/day, so even a tight 0.25° orb makes a "trigger" a 15-day window; Mars at a 1° orb is a ~4-day window; only the Moon (13.2°/day) can mark an hour (a 0.25° orb window lasts ≈ 54 minutes), and only event-moment angles carry true minute structure. A falsification of the Moon channel therefore required a new evidence class: events whose minute of occurrence is independently documented, on charts whose birth time is already known from a certificate. This is a strictly stronger test than recovery — no candidate scan, no multiplicity, each event is one clean test point.

The timed-events registry and its audit

Three frontier models were asked to assemble candidate subjects and times from primary sources. The audit that followed is part of the result: ChatGPT contributed 7 conservative, correctly-graded rows; Gemini contributed 14 rows that survived verification with AA→A downgrades where the "second source" was wire copy repeating a palace bulletin; DeepSeek's 19 rows were rejected wholesale — every load-bearing check failed (a "AA" Lincoln birth time that no record supports, JFK Jr. off by nearly fifteen hours, Elvis's death off by an hour, invented certificate numbers). Three further audit rounds — covering RFK, Jackie Onassis, Einstein's cremation and Princess Anne, among other items — applied the same standard: RFK and Onassis anchors failed the AA bar (family memory), Einstein's cremation is real but same-day (not an independent event), and Princess Anne qualified with two A-grade childbirth bulletins after her wedding's 11:30 was correctly excluded as retrospective-only (C-grade). The final registry holds twenty audited event rows — nineteen primary events across eight subjects plus one partial-subject row (Einstein, 1 event, below the ≥2-per-subject bar), each birth anchor AA, each event time AA/A with two independent sources, each timezone/DST offset verified for the event's own date (e.g. Paris in August 1997 is UTC+2, London in May 1981 is BST). Einstein's single event is retained in the registry but excluded from the pooled count under the spec's ≥2-events-per-subject rule. The audit itself required a self-correction: the initial row count was 14, which did not match the listed subjects (who sum to 17); the discrepancy was caught and corrected before any engine run, with the CSV as ground truth.

Channel, nulls and result

Each event contributes one test: is the transiting sidereal Moon within orb of a classically-grounded natal target degree at the exact documented UTC minute? Targets per type (whole-sign frame, cusp = sign boundary): death/accident → natal Moon, lagna, Saturn, Mars, Rāhu; marriage → natal Moon, lagna, Jupiter, Venus, 7th cusp; childbirth → natal Moon, lagna, Jupiter, 5th cusp. Primary orb 0.25°, sensitivity 0.5°/0.1°. Nulls: hour-shuffle (primary; the event time uniform over the 1,440 minutes of its own UTC day) and day+hour (date additionally drawn from the type's age-window pool), 20,000 replicates each plus an exact Poisson-binomial cross-check. M-8 (Moon in the 8th from natal Moon or lagna, sign-based) is reported secondarily.

Reading. At the pre-registered primary orb the verdict is rule 2 — anti-signal or indistinguishable: real events do not land on Moon–significator degree-conjunctions more often than the other 1,439 minutes of their own days (0 observed against a 0.07 null, p = 1.0; permutation and exact p-values agree to every reported digit). The power caveat is honest and stated: a single-target crossing occupies only ~2% of a day and occurs on only the ~17% of days the Moon's arc reaches the target, so 19 events is a modest test — but the direction admits no candidate-signal reading. Stage 2 of the spec (birth-time recovery from minute-dated events) is not entered.

What this study proves, and what it does not

• Raw gochara, day-granular: the classical trigger channels recover no birth time within ±30 minutes on 23 certificate-grade lives (0/23), and the cohort's apparent concentration (mean percentile 17.2; 4/23 in the top 5%, p = 0.026) does not beat a life-course-preserving null (p = 0.73). Candidate signal at most — not confirmed — and unlike the daśā layer it is not an anti-signal: real times rank better than chance-shaped expectations, which is why the layer was refined rather than closed after Stage 1.
• Aṣṭakavarga: the classical refinement, operationalised from the JHora contribution table with the standard sign thresholds, fails its gate on every arm and materially degrades a degree-based ranking (mean percentile 17.2 → 45.5; top-5% 4/23 → 1/23). The AV grid is a ~2-hour step function by construction; its apparent top-5% "hits" are plateau artefacts. AV may inform sign- or season-level judgement; it cannot resolve minutes, and adding it to a minute-seeking engine hurts.
• The hour-level Moon channel: zero of nineteen independently minute-dated events on AA-anchored known charts fall on a Moon–significator conjunction at the documented minute (null expects 0.07; p = 1.0 at every orb), and the sign-level M-8 channel is null (p = 0.90). The strongest transit evidence class the public record can supply shows no trigger signal at the only transit resolution that can mark an hour.
• Not proven — and important to say so. (a) This tests transits of the classical significator set against natal points from the lagna and Moon; it does not test every transit doctrine ever taught. (b) Event-moment angles (transits and directions to the angles as they cross the sky at ~1°/4 min) remain untested — they require minute-dated events at a known site, which no current certificate-grade registry provides; the v3 spec defers them (its "Ang" channel) rather than inventing coordinates. (c) Stage-3 power is modest by the spec's own admission (19 events), though the direction is unambiguous. (d) Nothing here touches the non-timing claims of astrology — personality matching and the lunar-phase market study are reported separately.
• The registry audit is itself a finding. Three frontier models asked for sourced, minute-dated events produced wildly different discipline: one response was retained in full, one retained with downgrades, one rejected wholesale for fabrication (including invented certificate numbers). Any study in this domain that sources its own evidence — ours included — must audit every row against the primary document. The full audit trail is published.

The resolution ladder, and the one rung left untested

Planetary motion bounds what each classical timing layer can ever claim. At the theoretical tightest orb (0.25°), Saturn moves ≈0.033°/day and makes a ~15-day window; Jupiter moves ≈0.083°/day and makes a ~6-day window; at the classical 1° orb these stretch to ~60 and ~24 days respectively. (Near a station, when either body is retrograde, instantaneous speed approaches zero and the actual window can be much longer — these are mean-motion best-cases.) Mars at 1° orb is ~4 days (≈0.52°/day); the Aṣṭakavarga grid is a ~2-hour sign-step; the Moon at 0.25° orb is a ≈54-minute window — the finest resolution any transiting planetary body can supply; and only event-moment angles move at ~1° per 4 minutes. Note: Stage 1 of this study used ±2° orbs (primary) and 1°/3° (sensitivity) — coarser than the 0.25°/1° figures above, which are theoretical best-case bounds. At the 2° orb actually tested, Saturn's window is ~120 days and Jupiter's ~48 days. The ladder figures characterise what each body can resolve, not what Stage 1 did resolve. Tested against that ladder: the daśā era (years) is falsified under two formalisations; raw gochara (days) is a candidate that fails its pooled nulls; Aṣṭakavarga (days, ~2-h step) fails and degrades; the Moon (hours) is null at every orb. The single untested rung is event-moment angles, which need a stronger evidence class than day-dated public biographies or even the minute-dated royal registry can supply (the event's exact minute and the event site's coordinates, so an angle to that site can be computed). The full ladder and its verdicts are drawn in the programme synthesis article (Classical Event Timing at Every Resolution).

Limitations

• Day-granular events for Stages 1–2. The same public-figure payloads as the daśā study — third-party event lists, type-labelled, dated to the day. These are the evidence a rectification customer would supply, but they are coarse relative to the slow-planet windows being tested.
• Stage-3 sample and power. 19 events across 8 subjects, every one audited to a primary source, is a genuinely rare and defensible dataset — and it is still small. The degree channel fires rarely even by chance (single-target crossings ≈ 2% of a day), so the spec's own power caveat applies; the finding rests on direction (0 vs 0.07) as much as on n. No formal power analysis is reported; the cohort was sized by availability of AA-grade, minute-dated events, not by a target effect size. For raw gochara (Stage 1), the pooled p = 0.73 cannot distinguish "no effect" from "underpowered to detect a real effect" — this is why the verdict is "candidate, not confirmed" rather than "falsified."
• Slow-body degree channels at event days were not re-tested at hour resolution. A Saturn conjunction cannot be resolved below ~15 days regardless of evidence quality, so Stage 3 tested only the Moon — the sole body whose motion supports hour claims. This is a physical bound, not a scope choice.
• Operationalisation. Fixed event→significator ontology, ±2° orbs (Stage 1) and 0.25° (Stage 3), whole-sign cusps, BPHS contribution table frozen at JHora defaults, Pushya Pakṣa frame. A different school's gochara (e.g. Varāhamihira's discrepant rules, all verified no-ops at JHora defaults) or a different target set could behave differently; we publish the exact encoding so it can be checked.
• Engine, not astrologer. A mechanical engine does not reproduce a practitioner's holistic judgment of a transit chart. It does, however, isolate the testable core of the doctrine — and it is the objective instrument a service would actually deploy.

Study timeline

Reproduction

The complete study is published with this article: the v1, Aṣṭakavarga and timed-event specs (committed before their runs), the 23-subject payloads and anchor audit shared with the daśā study, the timed-events registry with its four-round audit trail and the rejected AI responses kept verbatim, every run CSV and per-draw array, the engines (`RESEARCH/notebooks/rectify_gochara_study.py`, `rectify_gochara_av_v2.py`, `rectify_timed_events.py`), and the run notes — so that every number above traces to the exact code that produced it. All negatives published in full, per the standing rule of the ATRAC Institute research programme. The companion daśā article is here, and the programme-wide synthesis with the resolution ladder is here.

Published 2026-09-07. This study is scientific research on the measurable claims of a planned service, published by the ATRAC Institute's Sidereal Telemetry Lab. It is not a validation of any astrological product and nothing in it should be read as a recommendation to rely on any astrological service. The birth-time rectification and transit-timing services remain closed until a protocol-fixed accuracy gate is met.