How we score islands

The overall score is the weighted average of the six, rounded to one decimal. If a factor has no score, it is left out and the other weights are scaled up to make 100%. Crete shows how it works: weather 8.5, beaches 9, things to do 9.5, access 3.6, value 8.6 and crowds 6.4 give 8.2.

Each island gets a weather score for every month, from four parts:

• Daytime high, 30%. Full marks for an average high of about 81°F to 88°F (27–31°C). Below that the score falls by 1.2 points per °C (about 0.67 per °F); above it, by 1.5 points per °C, because heat past 88°F is harder on sightseeing than a cool day.
• Sea temperature, 25%. Zero at 64°F (18°C) or colder, full marks at 79°F (26°C) or warmer, in a straight line between.
• Rainfall, 35%. Full marks for 2 inches (50 mm) a month or less, zero at about 9.8 inches (250 mm) or more.
• Sunshine, 10%. Zero at 6 hours a day, full marks at 11 hours.

Then we subtract storm points for islands in a hurricane, typhoon or cyclone basin (next section), and keep the result between 0 and 10. The yearly weather factor is 60% the average of an island's five best months and 40% the average of all twelve. That rewards an island with a long, reliable good season without ignoring its bad months.

Storm points come from the official seasons of each basin, set by the NOAA National Hurricane Center for the Atlantic and Central Pacific and by the regional centers for the Western Pacific, South Pacific and southwest Indian Ocean. They reflect normal risk by month, not a forecast.

Aruba, Curaçao, Bonaire, Grenada and Barbados take half the Atlantic deduction, because they sit on the southern and eastern edge of the hurricane belt, where direct hits are rarer. The Mediterranean, the Canaries, Madeira, the Azores, Thailand, Bali, the Maldives, the Seychelles and Zanzibar take none; their wet seasons show up in the rainfall part instead. Our hurricane and cyclone guide sets out the seasons themselves.

Daytime highs, rainfall and sunshine come from the Open-Meteo Historical Weather API, averaged over the ten years 2016 to 2025 at the grid cell of each island's main airport, the one listed on its page. Open-Meteo's archive blends ECMWF reanalysis and forecast-model data such as ERA5 and ERA5-Land, so it is a gridded estimate, not a single weather station. Sunshine follows the WMO definition: hours when direct sunlight exceeds 120 watts per square meter.

Sea temperatures come from the Open-Meteo Marine Weather API, which serves a global sea surface temperature dataset from January 2022. We average every reading from 2022 to 2025 at the sea cell nearest the airport, or nearest the island's center where the airport cell had no data.

Using the airport keeps the method the same everywhere, but it has limits. On big islands the airport side may not be where you stay: the Big Island's numbers are for dry Kona, not wet Hilo. Every island page shows the full 12-month table, so you can check the months yourself.

These two factors carry half the weight, because they are what most people travel for. They use one scale for every island:

• Beaches: 10 means several excellent, swimmable sandy beaches near where visitors stay; 8, very good beaches with caveats such as seaweed, surf, rocky entries or a drive; 6, decent but limited or seasonal; 4, few swimmable beaches or mostly rock and pebble; 2, not a beach island.
• Things to do: 10 means a week or more of varied days, from nature and culture to food, water sports and day trips; 7, several good days beyond the beach; 5, the beach plus a handful of excursions; 3, mainly the resort.

The reason for each score is printed on the island's page. Scores move when the island does: Kauai's things-to-do score is 7 while Hurricane Lowell's closures keep Waimea Canyon and Koke'e shut, and Jamaica's is half a point lower while attractions reopen after Hurricane Melissa.

Access starts from the average great-circle distance between the island's main airport and four US hubs: New York JFK, Chicago O'Hare, Dallas/Fort Worth and Los Angeles. An average of 1,200 miles (1,930 km) or less scores 10 and 9,000 miles or more scores 2, on a log scale between. An island with at least one US airport flying nonstop gets one extra point, and half a point more with five or more, up to 10. Curaçao averages 2,516 miles from the four hubs and has six US nonstops: 8.6. Bali averages 9,583 miles with none: 2.

Value is the average of two hotel prices: the cheapest double rated 3 stars or better, for two adults, in the island's main visitor area, on Wednesday, January 13, 2027 and Wednesday, July 14, 2027, taxes and fees in, from Kayak (or Google Hotels or Booking.com where Kayak failed; each page says which). An average of $60 or less scores 10, $600 or more scores 1, on a log scale. Maui's $256 average scores 4.3.

Crowds divide annual overnight visitors, plus a third of annual cruise passengers, by the island's land area. A day ashore from a ship counts for less than a stay. About 260 visitors per square mile (100 per km²) or fewer scores 10, about 51,800 per square mile (20,000 per km²) or more scores 1. Santorini, with about 24,900 visitors per km² by this measure, scores 1.0. Molokai scores 10.

Thirteen islands have no crowd score, because no published figure measures visitors to that island for a full year. Rather than guess, we leave the factor out and weight the other five up. The gaps are:

• Key West: the only count covers domestic visitors to the whole Florida Keys.
• Nantucket, Hilton Head Island, St. Thomas and St. Croix: no official visitor count; the closest proxy is FAA passenger boardings at the airport, which include residents.
• St. John: only national park recreation visits are published.
• Tortola and Virgin Gorda: the latest figure covers January to September 2025 for the whole British Virgin Islands.
• Palawan: only a figure for the whole Philippines.
• Koh Lanta and Koh Phi Phi: only a figure for all of Krabi province, mainland included.
• Bora Bora: only a figure for all of French Polynesia.
• St. Barts: only total air and sea arrivals, which is not a stayover count.

Where a country figure covers essentially the islands visitors go to, as in the Maldives, the Seychelles, Fiji and Jamaica, we use it and say so on the island's page.

Every island page ends in a verdict: worth it, conditional or skip. A conditional verdict takes 0.4 points off the overall score and a skip takes 1.0, so an island that suits only some travelers cannot outrank one with the same numbers that suits almost everyone.

• World and region lists sort by the overall score.
• Month lists score each island as 55% that month's weather and 45% its other five factors, verdict adjustment included. An island needs a weather score of 7 or more that month, and no region takes more than three places.
• Traveler lists (kids, teens, couples, solo, family) include only islands whose page tags them for that traveler, sorted by overall score. The no-passport list keeps US islands only. The dog list keeps islands where bringing a dog is easy or possible, easy ones first.
• Budget lists use 60% value and 40% overall, among islands scoring 6.5 or more overall.
• Region lists for a traveler add 1.5 points to tagged islands, so they come first but the others still show. Winter-sun lists use 60% the average December-to-February weather and 40% the other factors.

See the results on our world ranking and a month list such as October.

The storm deduction is long-run risk, not this year's forecast or damage. Damage enters through the editors' scores and the verdict, as with Kauai, Jamaica and the Big Island in 2026. Hotel prices are two dates, not a yearly average. Visitor counts are the latest full year we could source. Air fares are not scored, because they change daily. When the data changes, the rankings change with it, and every source we cite shows the date it was checked.