How we pick the top 10 lists and books

Each activity guide on Park Atlas has a ranked list of the best places for that activity and a short shelf of books. Neither is our opinion. The rankings come from lists other people have published, and the books come from the stores and publishers that know each park best. This page explains how both are built and checked.

Where the top 10 comes from

For each activity we collect between 10 and 20 published lists of the best places for it. A list counts only if it names specific places and was published or updated in the last six years or so. We sort the lists into four kinds of source. The first is official and specialist groups, such as the National Park Service, park conservancies and national associations for the activity. The second is national magazines and newspapers. The third is guidebook authors and expert writers with real experience of the places. The last is reader-driven sites such as trip review and trail apps, which we count as a sign of demand and weight lowest.

We record the places each list names and the order it gives them. We copy names and ranks only, never another publisher's descriptions. Each place is then matched to its record on Park Atlas. A match we are unsure of is left out of the scoring rather than guessed.

How the lists become one ranking

A place earns points from every list that includes it. A higher spot on a list earns more, and a spot on a short list counts for more than the same spot on a long one. Points are weighted by the kind of source and by how recent the list is, and when one publisher has several lists they share one publisher's weight. The top 10 is the ten places with the most points. We allow only two adjustments, and both are logged. When a list mixes kinds of places, such as trails and parks, we rank each kind separately. And when two places are within 5 percent of each other, the one on more lists goes first.

Each pick carries a label for how widely the lists agree. A lock appears on at least half of the lists. A strong pick appears on at least 30 percent of them. A candidate appears on fewer. We also show how many of the lists include each place.

Our own data never changes the order. The number of things-to-do pages the Park Service publishes for a park, or how many people visit it, says nothing about whether it is a good place for the activity. When an activity has too few published rankings to build a consensus, the guide shows researched highlights instead of a ranked list and says so.

Checking the picks

Before a list goes live, we check each pick against the Park Service's own pages for closures, permits, lotteries and seasons. A pick stays on the list either way, and the note under it gives the fact that matters, such as a permit you need before you go. A place closed for more than a year is left off.

Every sentence written about a pick cites a source. A script confirms that the quoted passage is really on the source page, and a second model judges whether the passage supports the sentence. A separate reviewer who did not build the list then re-runs the scoring from the saved lists, spot-checks list entries against the original pages, and checks each sentence against its source. Anything that fails is fixed or removed.

Books

A book shelf holds three to five titles. We look first at the bookstores run by park associations and conservancies, because a book sold in a park's own store was chosen by people who know the place. We also use Park Service reading lists, the standard guidebook and field guide publishers for the activity, and book awards. Each candidate is scored on how well it fits the page, how useful it is on a trip, where it was recommended, how current the edition is, the author's authority on the subject, and whether it adds something the other picks do not. Only books that score well make the shelf.

Every pick is checked by hand on Bookshop.org to confirm the edition is the one we name and that it can be ordered. A reviewer who did not pick the books then checks each one against its evidence. A book that fails is dropped, and a guide with fewer than three books that pass has no shelf at all. We earn a small commission on books bought through these links. It does not change what we list.

The evidence

The lists, the scoring output, the matching decisions and the reviewers' verdicts for every guide are kept with the site's data, so each ranking can be reproduced. We refresh the lists every year, or sooner when the published lists change.