Day-level acceptance bands (CSV-only audit targets)
These targets define a simple, reproducible bar for CSV-only audits. They do not change the evaluator; they guide what “good enough” looks like for day-level transits.
CSV integrity (structural)
- Header/coverage/uniqueness checks: PASS for clean daily grids.
- Nodes identity: exact
Ketu(t) = wrap360( Rahu(t) + 180 )→ PASS when frames match.
Events & angles (typical bounds)
- Rāśi crossings (days): for bodies with enough events in a 5-year TEST window, target
rasi_cross_MAE_days <= ~2.0(slow outers up to~2.2). - Cusps (deg):
cusp_dist_MAE_deg <= ~3.0for slow outers; Jupiter/Saturn often tighter. - Stations (days): informative but advisory; slow outers naturally show larger timing MAE.
- Degree summaries (optional):
MAE_deg,P90_deg,MAX_degare reported but not used as hard gates for event fidelity.
Sparse-event policy (slow outers)
When a 5-year TEST window yields < 3 matched crossings (e.g., Uranus), extend the window (e.g., 40y) and/or rely on cusp/station metrics. This behavior is by design and should be documented alongside the metrics table.
Notes for auditors
- Keep frame/time fixed: Lahiri sidereal, one daily timestamp, degrees in
[0,360)with one-wrap display viawrap360(x). - Pairing caps: crossings
±60 d, stations±90 d. - Use
NaNwhen an event metric is undefined (no matched pairs). - Report the exact TEST span, timestamp, and pairing caps with your results.
Navigation
Back: SSM-JTK – Results to Date — Per-body highlights (3.3)
Next: SSM-JTK – Results to Date — Artifacts & manifest (3.5)
SHUNYAYA 