Purpose
A simple, repeatable way for readers to audit results using a reputable public ephemeris. This uses your local CSVs or ephemeris outputs plus the deterministic metrics below. It does not change the evaluator in any way.
Metrics (ASCII)
- Phase error on
[0,30)
Letd = ((L_true % 30) - (L_pred % 30) + 15) % 30 - 15.phase_err_deg = |d|
Aggregates:phase_MAE_deg = mean(phase_err_deg),phase_P95_deg = P95(phase_err_deg). - Rāśi crossing timing (days; positive = late)
rasi_cross_err_days = (t_cross_pred - t_cross_true)
Aggregate:rasi_cross_MAE_days = mean(|rasi_cross_err_days|)after nearest-neighbor pairing within±60 d. - Station date timing (days)
station_date_MAE_days = mean(|t_station_pred - t_station_true|)after nearest-neighbor pairing within±90 d.
Procedure (repeatable)
- Fix a span (e.g.,
2000-01-01..2099-12-31) at one daily timestamp (e.g.,05:30 ISTor00:00 UTC). - Compute model
L_pred_degusing the one-line evaluator from a manifest (same timestamp). - Obtain reference
L_true_degfrom a reputable public ephemeris using the same frame/time: geocentric, sidereal Lahiri, same daily timestamp. - Compute metrics above per body and compare with the acceptance bands in §3.4.
Notes & tips
- Use one-wrap display for comparison:
wrap360(x) = x - 360*floor(x/360). - Always work on unwrapped tracks when interpolating events; wrap only for daily angle display.
- Document the exact span, timestamp, and pairing caps used in your report.
- When events are sparse (e.g., Uranus in short windows), expect
NaNfor crossing timing; rely on cusp/station metrics.
Lightweight reporting template
Span=YYYY-MM-DD..YYYY-MM-DD Timestamp=<e.g., 05:30 IST> Frame=Lahiri (geocentric)
Body phase_MAE phase_P95 rasi_cross_MAE_d station_MAE_d
---- --------- --------- ---------------- -------------
...
Navigation
Back: SSM-JTK – Results to Date — Alignment utility (3.6)
Next: SSM-JTK – Locked Kernels – Tiny Reproduction Kit (4)