# Hunt H-0001: Sun-like stars that Gaia caught fading

**Question.** Which main-sequence FGK stars did Gaia see dim (2014-2017), and which of those has nobody
explained?

**Funnel** (every step is a script in `scripts/hunt_h0001*.py`; every number below is reproducible):

| step | stars | what removed them |
|---|---|---|
| Gaia DR3 `vari_summary`: G 10.5-15.5, FGK colour, near the main sequence, parallax/error > 10, RUWE < 1.4, clean images, G range > 0.2 mag, skewness > 2.5 (fading) | 22,125 | |
| not classified by Gaia as an eclipsing binary | 2,020 | 20,105 eclipsing binaries |
| a dip deeper than 0.8 mag confirmed on 2+ transits, excluding 3+ mag single-transit window artifacts | 18 | eclipses of similar stars can't exceed ~0.75 mag; 62 stars carried spurious transits |
| not already explained in VSX, SIMBAD or the ZTF variable catalogues | 3 | 8 known Algol-type binaries (ASAS-SN, ZTF, SSS, OGLE), 1 known young star, 1 blend with a catalogued eclipsing neighbour; 6 more carry only a Gaia "VAR" entry with a pipeline period |
| followed up with Oddsky cases | 3 | NSV 1302 (OS-0010): a known Per OB2 young star, characterised as an irregular dipper. Gaia DR3 2037357495383317504 (OS-0011): a new Algol-type binary, P = 2.52961 d. **Gaia DR3 4482344869315388928 (OS-0012): a new eclipsing system, P = 2.380206 d, whose Sun-sized occulter shows no light of its own** |

The two binaries were missed for a simple reason: Gaia's pipeline gave them periods of 0.037 and 0.025 days, VSX
inherited those, and nobody looked again. The `eclipse_period` tool (phase-clustering of the rare deep points,
across instruments) found the real periods in seconds, and Gaia's 2014-17 dips fall exactly on the ZTF ephemerides.

**Verdict.** No new main-sequence dipper in this sample. The negative result is the finding:
with ~50 sparse Gaia epochs, deep "fading" on a Sun-like star is almost always an Algol (a hot dwarf
eclipsed by a cool subgiant) or a Gaia windowing artifact. Two lessons now built into Oddsky:
Gaia epoch photometry is a series (`GAIA:G/BP/RP`), and the Skeptic's blend check caught the one
case where the dips belonged to a catalogued neighbour.

Files: `query.adql`, `candidates.csv` (22,125), `scored.csv` (2,020), `vet/*.json` (external checks),
`epoch/` (Gaia light curves, not committed: re-fetch with the scripts).