Read it by register, not by row
Every card looks the same until you know its register. A deterministic physics formula, a live data feed, an observation that only annotates, and a hard safety gate carry very different weight. The badge on each card tells you which one you are reading. Open a card for the structured detail: inputs and units, output, the activities that use it, its role in the verdict, the sources behind it, its limits, and the reference.
| Signal | Role | Register |
|---|---|---|
| Wind | Primary suitability | Production physics |
| Gust factor | Confidence | Production physics |
| Waves | Suitability | Production physics |
| Tide | Timing window | Production physics |
| Lightning | Hard safety gate | Hard safety gate |
| METAR / buoy | Validation | Observation validation |
Every coefficient, by register
01Foundational
02Marine
03Atmospheric
04Comfort
05Hydrology
06Cryosphere
07Cross-domain
08Hazards
09Health
10Engine
Four ways a number can be true
Data and validation cards carry an honest provenance label. A measured sensor reading, a satellite retrieval, a model forecast and a historical archive are not the same kind of truth, so we never present them as one. Where a representative and fresh observation exists near a spot, it corrects or annotates the forecast. Where it does not, the forecast stands on its own and confidence reflects that.
Confidence on a score reflects upstream forecast skill, agreement between providers and the maturity of the activity profile. A validation card never silently overrides a forecast; it annotates it with an observed value and its provenance.
How we check ourselves
The engine is graded in the open. Three standing pages hold the record: what the scores got right, how good the upstream forecasts are, and whether that skill is drifting over time.
Where the scoring record is graded: Brier Skill Score, CRPS and reliability over every audit-logged score, with a public CC BY 4.0 export.
Per-cell provider skill against ERA5 reanalysis. The consensus blends providers by inverse-variance pooling per region, variable and horizon, then METAR, sea-surface temperature, wave and real strikes check the blend against observed reality.
The Concept-Drift Guardian: a statistical-process-control monitor over per-cell forecast skill, with charter documents mapped for parametric MGAs and regulators.
A research track, clearly labelled
Inverse Suitability (IRT)
ResearchA research track exploring how continuous physical conditions and reported outcomes can separate condition difficulty from user skill.
What this is not
Goable is decision support. It does not replace local safety procedures, official warnings, instructor judgment or activity-specific regulatory requirements.
Forecasts carry uncertainty. A confident verdict is a probability, not a guarantee.
Observation validation only applies where a representative and fresh source is near the spot. Elsewhere the forecast stands alone.
Safety gates are conservative. When a safety feed is missing, the engine flags the gate as unevaluated rather than passing it silently.
The page covers activities with real hazard, including lightning, rip currents, rockfall, avalanche terrain and aviation-adjacent conditions. Treat the verdict as one input among your own checks.