Methodology

Three scores

• SkyScore: how good a point is for seeing a specific event.
• StayScore: how well a place to stay suits astronomy, based on its facilities.
• AstroScore: 70% SkyScore + 30% StayScore.

Rules we don't break

• If an event physically can't be seen from a point, SkyScore and AstroScore are 0. Facilities don't make up for it.
• Partial phases score on a harsh scale: 99% is nothing like totality.
• In a solar eclipse, light pollution has zero weight.
• A factor with no data is never made up: it is left out, its weight is redistributed and the confidence we show goes down.
• Months ahead, we use historical climatology, labelled as such. We only give forecasts within the model's range.
• Any future sponsorship will never change these scores, and will be marked separately.

StayScore weights

• Unrestricted night access: 20%
• Telescope available: 15%
• Astronomy guide: 15%
• Terrace or garden with a clear horizon: 12%
• Outdoor lights can be turned off: 12%
• On-site or nearby observatory: 10%
• Outdoor power sockets: 8%
• Designated dark area: 8%

An unconfirmed facility earns 50% of its weight.

What a place states on its own website (telescope, observatory, astronomy activities, reduced lighting, pitches with power or 24-hour reception) counts as an unconfirmed facility: we review it by hand but don't check it on site. Activities count as an astronomy guide, and 24-hour reception as night access.

How places to stay are ordered

We sort with a provisional score: 70% location and 30% facilities and available data.

• Location: in a central eclipse, from 50 to 100 depending on how close it is to the centre of the path. Outside the path it drops quickly below 25.
• With no path, location barely changes across the area we cover, so only facilities and data count.
• Facilities and data: half comes from the data we know (Starlight certification 50, own website 40, Wi-Fi 20, accessibility 20, parking 10 and pets 10, up to 100) and the other half from the StayScore. With no known astronomy facilities, that half is 0 for everyone alike. Certification never counts as location.
• Facility icons show what each place publishes on its website, and they count towards the StayScore at half points.
• No commission or sponsorship changes this order.

How to read “Why this score?”

• Each factor is worth 0 to 1. It shows as “in favour” from 0.6, “against” below 0.4 and “average” in between. These are fixed thresholds, not an opinion.
• A factor without data neither adds nor subtracts: its weight is redistributed and the confidence we show goes down.
• Each explanation states the model version used to calculate it.

Evidence labels

• Verified: from an official source or checked by us (for example, NASA's eclipse path).
• Calculated: the result of our calculations with ephemerides and published data.
• Reported by the community: contributed by users and reviewed before publication, without a site visit.
• Unverified: from an external source we haven't checked on the ground.
• Satellite: observed from space. It isn't a measurement of the sky from the ground.

Community spot ranking

Each published spot has a score from 0 to 100. It's a community score: it never feeds into an event's SkyScore.

• 50% dark sky: when a spot is published we measure the night lights seen by satellite (NASA Black Marble 2016) within about 6 km, giving more weight to what is closest. The index goes from 0 (urban sky) to 1 (no lights nearby). It's an estimate from satellite radiance, not an SQM measurement or a Bortle class.
• 50% ratings: the average of sky and experience (1 to 5) with 3 imaginary ratings of 3. That way a single 5 doesn't beat ten ratings of 4.8.
• If the night-lights source doesn't respond when a spot is published, it counts only its ratings, and this is shown.
• Nobody can set the index of their own spot: the server calculates it on approval.

Spot locations and sensitive places

• Whoever adds a spot chooses whether the exact point or only the area (rounded to about 1 km) is published. The area is all that the map, the public API, structured data and share images see.
• If someone marks a place as fragile or protected, it is published as approximate at most, and moderators can generalise it further or not publish it.
• Moderators can also generalise the location of any published spot after a report.
• We don't publish private land or areas closed at night. "Perceived darkness" is the contributor's impression and is shown as such: it is not a measurement or a Bortle class.

Known limits

• Local duration is calculated with a circular lunar limb. Compared with IGN values, the measured differences range from 0.1 s (Cádiz) to 5.6 s (Melilla).
• Because of the same model, the duration calculated at the best points in Egypt (about 6 min 26 s) exceeds NASA's official maximum (6 min 23 s) by a few seconds. We show the official figure as the reference and local ones as our own calculation.
• The local horizon isn't measured yet, and terrain isn't taken into account for visibility.
• Historical cloud cover has a resolution of 1° and is a monthly average across all hours of the day.
• There is no point-by-point light pollution measurement yet. The map's night lights show satellite radiance, not SQM readings or Bortle classes.
• Towns are scored at their centre point: results can vary within the same city.