The data. Waves, swell height, period and direction come from Open-Meteo's marine models (ECMWF WAM, GWAM, Météo-France and NCEP WaveWatch), blended together. Wind, gusts and pressure come from six forecast models (ECMWF, GFS, ICON, GEM, Météo-France and UKMO) plus ECMWF's 51-member ensemble. All of it is hourly, 7 days out. Morning = the 09:00–13:00 hours averaged, afternoon = 13:00–18:00. The page re-fetches every hour on its own.

The percentages. Each one is the chance that activity is closed or cancelled in that time slot. Every activity has its own breaking point: the Blue Grotto's mouth is barely a metre high, so it closes once swell lifts the entrance water 30–40 cm. It faces northwest, so N/NW/WNW swell or wind hits it head-on while a southerly barely reaches it. Boat tours give up around 1.2 m or 18 kt. Hydrofoils cancel before car ferries; Amalfi Coast routes before both.

The verdict chips. Calm (under 13% chance sea activities are cancelled), Good (to 30%), Uncertain (to 50%), Likely off (to 72%), Off (above). The figure weighs the Blue Grotto at 62% and a boat tour at 38%, since those are the plans that break first.

The confidence line. Starts high and drops with each day further out, when the forecast models and the ensemble disagree on wind, when pressure is falling fast (a system moving in), and when conditions sit right on a threshold. Pulling in six models plus a 51-member ensemble means disagreement is measured across many independent forecasts, not just two. It's about the forecast, never about how open something is.

The written summaries. Generated from the numbers by fixed rules, so the wording changes as conditions do: grotto odds above 80% closed produce a "shut" line, 55–80% a "boatmen often refuse" line, 30–55% a "borderline" line, and below that a calm line or a wind warning for boat trips. A note is appended when the afternoon is meaningfully better or worse than the morning.