How baromoon works

deterministic piping · cited facts · no AI in the report path

The short version: baromoon reads the weather forecast, a 75-day weather archive, the sun and moon's positions, and a cited tackle library — then scores every candidate window and tells you when to be on the water and what to throw. Same inputs, same report, every time.

What goes in

• Weather — Open-Meteo forecast (16-day cap) and archive (75 days): temperature, wind, cloud, precipitation, barometric pressure and its trend.
• Water — a first-order water-temperature estimate from the air-temp history (lag ≈ 4 days), plus a warming/cooling trend. A USGS gauge is used when one is live within 30 km of the water; no registry lake currently has one, so the estimate is labelled as such.
• Sky — Swiss Ephemeris: sunrise, sunset, civil dusk, moonrise and moonset, moon phase, solunar majors (moon overhead/underfoot) and minors (rise/set), planetary hours, and — for a profile — natal transits.
• Tackle library — every lure, rig, bait, knot, and line entry carries cited provenance, quoted from manufacturer catalogs (T1) or state-agency guides (T2).

What comes out

Each candidate window gets a score and a rank; the report shows the prime moment, the top picks with a cited why, and what you're missing that scores better for these conditions. A capped tie-break layer adjusts for declared cover, season, spawn phase, and a derived fish-position state — one raw fit is 0.25 points, each dimension caps at ±0.5, and the whole layer caps at ±1.0. Signal magnitudes come from cited sources where one exists and are labelled T5 (editorial) when it doesn't.

What doesn't score

Multi-model forecast agreement (GFS/ECMWF/ICON spread), creator and tournament trend signals, and the tackle offers are display-only. They are shown after the ranking is set and never feed back into it.

The calibration ledger

Every logged trip — catches and skunks alike — is replayed blind: the model sees the logbook only as of each session, and the weather comes from the archive. As of late September 2026 the ledger held 13 sessions: 6 exact, 3 style, 2 miss on presentation; weighted by catch, 37% exact / 60% style / 3% miss, with score separation +0.4. When the data disagrees with the model, the model changes; the scoring rules live in the tackle library's condition contract.

Money

Tackle links are added after the picks are set, always disclosed, and never influence the order. Manufacturer pages carry no commission; affiliate links do, and are marked. See the affiliate disclosure. baromoon is free to use — no account, no paywall.

What it isn't

Conditions are estimates — a forecast, an inferred lake state, and an air-temp water model — so a report is a plan, not a promise. Access rules, licenses, and regulations are yours to manage; baromoon doesn't enforce them. And no AI writes the facts at runtime: the numbers are computed from the sources, every time.

Looking for the machine-readable side? The API and embed expose the same deterministic rankings.