v19-web · 260924 1850 ETGreece 2027research mode17 pairs

Seventeen Island Pairs

Josh's shortlist plus five the list missed. Every pair is one town stop and one resort stop. Rate them, mark them in or out, and leave notes. Nothing here is a selection — this is the set still worth arguing about. ← back to the trip page

TripJosh's stated criteria, answered 9 Sep. Rank out of 63; lower is better.

GirlsWhat Elsa and Evie would actually use. Rank out of 63; lower is better. A weak girls’ rank is a factor, not a filter (Josh, 13 Sep) — it never removes a pair that is otherwise strong.

ContrastWeighted out of 13. Landscape and architecture count triple — they are what make somewhere feel like a different place.

CrossingHow you get between the two stops, and how often the boat or plane really runs.

LodgingWhether a property has been researched at twelve people. Most have not.

LookPhotos are Wikimedia-licensed. Map, walk-through video and topic links open in a new tab.

StayGreen links go straight to the properties found for that base.

WindClick to open. Traveller reports and Greek-language sources, swept 13 Sep. The meltemi blows from the north, so the question is always which side of the island you are on.

Your ratings and notes save in this browser only. They are private to you and are not sent anywhere. Use Copy my notes below to send them to Josh.

View Every pair is two stops, so places repeat. Switch to see each one once, grouped by role.

Sort

The wind, actually measured

Pulled 13 Sep from Windfinder's airport weather stations — real observations, roughly 2000 to 2026, not traveller opinion and not a windsurfing school. Average wind speed and gusts in knots. This is the first instrument-grade, month-by-month comparison in the project.

Base	Station	June avg	June gust	July avg	July gust	vs calmest
Kos	Kos Airport	10	24	12	26	1.0×
Crete — Elounda	Heraklion Airport (proxy, 70 km away)	10	17	12	19	1.0×
Syros	Syros Airport	10	22	12	26	1.0×
Naxos	Naxos Airport	11	21	13	25	1.1×
Rhodes	Rhodes Airport	12	18	13	18	1.2×
Astypalea	Astypalaia	12	19	14	22	1.2×
Karpathos	Karpathos Airport	18	34	22	37	1.8×

Six of the seven sit within two knots of each other in June. Karpathos sits eight knots above the pack. Its June gusts beat everybody else's July gusts, and its June average beats everybody else's July average.

This is the finding the traveller reports could not deliver, and it survives the objection I raised against my own earlier evidence: these are civil airports, not windsurfing schools.

Two station-location biases, and they point opposite ways. Rhodes airport is at the island's north-west tip — the end travellers call windy — and it still posts the lowest July gusts on the board, so Pefkos on the sheltered south-east should be calmer than the number shown. Karpathos airport is at Afiartis on the southern tip, beside the windsurf beaches, so Pigadia and Amoopi will read lower than 18. But the gap is wide enough that even a generous correction leaves Karpathos the windiest candidate.

Not in the table: Chania and Kalymnos. Windfinder's Chania station reports June at 5 knots with gusts of 9, which is implausible for Crete's north coast and is not an airport reading — ranking it against the others would be false precision. Kalymnos has no station.

The five small Dodecanese islands

Symi, Nisyros, Tilos, Chalki and Kastellorizo were all in the research already, and all five were ruled out on 10 September in a single line: they “fail on resort-grade beach infrastructure, on access from Athens, or on being too small.” That test was wrong for the job they would actually do. They were judged as places that had to supply the beach holiday. As a three-night town stop, with the beach coming from the other island, the objection mostly evaporates.

Airports: only Kastellorizo has one. Symi, Nisyros, Tilos and Chalki are ferry-only, confirmed island by island. Kastellorizo has KZS, served by Olympic Air from Rhodes alone, 40 minutes, six days a week. No seaplane service operates to any of the five — the licences are announcements, and as of February 2026 the operator held only a charter licence.

The connections are real, and better than the rest of this board. But there is a trap worth naming: a day-trip excursion is not a crossing. Dodekanisos Seaways markets day trips from Rhodes to seven islands, and the Kardamena “Nisyros boats” sold from south-coast Kos hotels are sightseeing round trips — one scanner reports zero scheduled ferry service on that route. The scheduled ferries leave from Kos Town. Book the wrong one and the plan strands.

Symi — on the board, as pairs S1 and S2. Daily to Rhodes, 45 to 50 minutes, twenty-plus sailings a week across four or five operators; six or seven days a week to Kos. The only one where twelve people could move on any date they liked. Costs: a hotel block rather than a house, and serious heat.
Chalki — on the board, as pair S3. Two independent Rhodes links, including a local ferry up to three times a day from Kamiros Skala. Four harbourside houses at four guests each is the cleanest lodging shape found anywhere. The Kos link is one operator twice a week and is not dependable.
Nisyros — viable, not added. The easiest access of all five, roughly ten scheduled sailings a week from Kos Town, eight of them on conventional ferries, fastest 50 minutes. The volcanic caldera you walk into is the best single half-day any of these islands offers a child. But no Nisyros property publishes how many people any room sleeps, so twelve cannot be confirmed from any source, and the crater is a half-hour drive from the only walkable town.
Tilos — viable, not added. Livadia Beach Apartments publishes six suites at four people each, so twelve is confirmed. About six sailings a week to each of Rhodes and Kos, and uniquely no day-trip market at all, so every sailing found is a real crossing. The problem is shape: the lodging is in Livadia, the capital and the medical office are in Megalo Chorio, and the only public transport between them is one bus, 25 to 35 minutes. Half the island's roads are sand and hire cars are not insured on them.
Kastellorizo — ruled out. The air link looks good on paper and the Blue Cave is genuinely the best ninety minutes on offer here. But it is a small turboprop into a light-aircraft airfield, the ferry to Rhodes runs three days a week on the only source giving real 2026 times, the Kos link is six to ten hours, there is no beach of any kind, no confirmed medical facility and no natural drinking water on a three-kilometre island. A deliberate week for two adults; not a hinge in a family itinerary.

One correction to carry forward. The widely repeated claim that Tilos was “Europe's first fully renewable-powered island” does not survive checking. The EU federation of regional energy agencies calls it one of Europe's first hybrid renewable microgrids, not the first overall, first only within Greece, and says it reaches full renewable coverage in the low-demand winter months — not in a July peak. What is defensible is the island-wide Natura 2000 designation holding about 10% of the world's Eleonora's Falcons, and zero waste since 2022.

What this page got wrong, and how we found out

On 14 September Josh asked how to catch coverage misses he had not personally spotted. A 126-file audit of the research folder followed, testing the 27 assertions this page makes. Six of them were false. Every correction below is marked on the card it belongs to.

The error was always the same shape: a conclusion about a whole, drawn from evidence about a part that was chosen by convenience rather than by design. An island judged from one coast. A corpus judged from memory. A market judged from one listing.

The six: “no 2027 rate exists” (live bookable June-2027 hotel rates, ferry tickets, flights and six villa tariffs were already on file); “the only property taking twelve in one house” (at least eleven others, and the Pefkos split claim has no source at all); “no teen programme at Kos” (Ikos Aria's Just4Teens, confirmed four ways); Karpathos wind “unknown” (it is contested, with two sources naming opposite coasts); “Astypalea solved” (five omitted disqualifiers); and the Chania–Rhodes flight, recorded at high confidence as running August to September only.

The standing fix is now written into the operating rules: state the search space before the superlative. Not “Rhodes fails the release-valve test” but “the six east-coast resorts researched in August all sit 40 to 50 km from the Old Town.” The second version cannot quietly become the first.

Where twelve people actually sleep

Researched 10–14 Sep, then audited against the full research folder on 14 September. Cards carrying a ⚠ CORRECTED block state something this page previously got wrong. Every figure comes from an operator or listing page that was actually opened, and where two sources disagree both are shown rather than one being picked.

Everything in the corpus was sized for six people, two families of three. The trip is now twelve people in four families. That is why the lodging evidence looks absent when it is really scoped to the wrong group.

One price exists on this whole board: Plaka Collection on Naxos, €300 per villa per night for 1 June to 20 July. Everything else needs a direct quote.

Three things the pairs reveal

The resort end has already converged. Eight of the nine finish on Rhodes, Kos or Crete. Those are the only bases with the scale to absorb a longer stay, and the only ones where a fifteen-year-old can leave the property and reach something.

The town end is genuinely open. Ermoupoli, Naxos Chora, Pothia, Chania old town and Astypalea Chora are five very different answers to the same question, and the research does not settle between them. This is where taste decides, not more data.

Contrast and connection trade off directly. Every high-contrast pair is cross-cluster, and every cross-cluster link runs one to four times a week. The same geography that makes the Dodecanese feel unlike the Cyclades is what stops there being a daily boat.

The price of a trip that feels like two places is a thin, weekday-pinned crossing.

Athens in the middle — ruled out

Five pairs on this page have no direct island-to-island link: Naxos + Kos (2), Chania + Kos (3), and candidates R2, R3 and R4. Each of them can only be joined by flying back to Athens and out again.

The open question used to be whether Athens could sit in the middle of the trip rather than at the start, which would have made those connections free — you were going to Athens anyway.

Josh decided against it on 13 September. Athens stays at one end. So those five are now labelled for what they actually require, via Athens, rather than showing a long ferry as though it were the route.

They are not deleted. A pair that needs an Athens connection is still buildable if the group ever accepts a backtrack day — but nothing on this page assumes they will.

The short-stop idea, recorded

Josh, 10 Sep: "maybe we spend 2 days and 3 nights in a place if it's very slow?" — and separately, that 8–9 nights at the long stop is too much. Live and unsettled. Three consequences if it holds:

The release valve becomes a property of the long stop only. Three nights is not long enough for two fifteen-year-olds to grind each other down, so small quiet towns reopen as the short leg. This is what keeps Astypalea and Pothia on the list at all.
A short stop needs a better crossing than a long one, because half a day goes each way and losing a day out of three nights is proportionally brutal. Kalymnos + Kos at 35 minutes gains most; Syros + Crete gains least.
A short stop must be walkable from the bed. Pothia, Astypalea Chora and Ermoupoli pass. Leros fails — four separate settlements, commerce in one and charm in another, and no car-rental evidence for the island at all. That is why Leros is absent despite ranking 3rd and 6th on the trip criteria.

What would actually move this list

A quote from Pefki Sea View Villas. The only property found anywhere that takes all twelve in one house and splits into two independent halves. Availability and price for the window are unknown.
A Naxos Chora property at twelve. Chora is the town stop in the two best-surviving pairs (1 and R1) and nothing is named there at any size.
A Naxos Chora property. Chora is the town stop in four pairs, including the two best-scoring ones, and nothing is named there at any size.
What actually went wrong at Talloires — situational (bored, stuck, no escape) or ongoing between the two girls. Different answers imply different trips, and only the first is solvable by choosing an island. One question to Kristin.
Cycladic lodging. Tinos, Sifnos, Amorgos and Andros have no group-lodging research at all. The Dodecanese is winning partly because it is the branch that got researched.

The constraints all of these sit inside

Twelve people, four families. Kids roughly 9, 10, 11 and 15 at travel time.
Three stops maximum — Ben and Phil do not want to move more than that.
Athens is one of the three and is already the biggest contrast in the itinerary, so the two islands must differ from each other or the trip is effectively one island with a ferry in the middle.
2027, corrected 14 Sep. This page said no 2027 rate existed. It does. Air is fully on sale and was verified to a live cart. Alimounda Mare on Karpathos was returning bookable, free-cancellation June-2027 room rates in August — €1,605 for five nights, 24–29 June 2027. Dodekanisos Seaways was selling Rhodes–Leros for 28 June 2027 at €40, every day of June selectable, and Rhodes–Kalymnos at €37. Six villas publish 2027 tariffs outright. A live 10 September search returned 60 Naxos properties priced for 26 June – 3 July 2027, including villas sleeping 12 to 16. What genuinely does not exist is a published 2027 ferry timetable, and 2027 rates for hotel-category properties are still mostly unpublished.

Ranks from the 63-pair matrix scored 10 Sep 2026. Girls' lens from the Talloires and girls' design notes of the same date. Crossing times and frequencies are 2026-season observations, not 2027 schedules. Photos are the Wikimedia-licensed set already in the trip repo, each one checked against its label before use; the Chania street shot and the Kos Zia shot are captioned loosely because the subject could not be confirmed from the image alone. Map links use bare place names and were spot-checked in a browser. Lodging researched 10–14 Sep from operator and listing pages. Corrected 14 Sep against a 126-file audit of the research folder that found 120 material coverage errors; every correction is marked on the card it belongs to. Earlier single-model research has now been cross-checked against the three-model reconciled files.