elsewhere

the rules

one surprising fact, one pin, once a day. find the place.

how a round works

  1. everyone gets the same challenge for the same utc date.
  2. read the fact. if you need help, reveal up to three clues, one at a time. each one narrows things down.
  3. drop a pin by tapping the map, pinning the crosshair, or searching for a place by name. drag or re-tap to move it. nothing is final yet.
  4. lock in. you get one official guess a day, and you won’t see any distance feedback until it’s locked, so there’s no hot-and-cold triangulating.
  5. see the answer, the story behind it, its sources, and how your score was built.

scoring, out of 1,000

three parts. accuracy matters most, clues matter more than speed, and every part is capped.

accuracy = 800 × A clue bonus = 150 × (1 − clues used ÷ 3) × A speed bonus = 50 × S × A d = distance beyond the target's radius (0 if you're inside it) A = 0.7 × e^(−d ÷ 2000 km) + 0.3 × e^(−d ÷ f) f = target radius × 20, kept between 25 and 250 km S = 1 under 30s, falling evenly to 0 at 5 minutes

A is a smooth curve with no brackets: being in the right part of the world earns most of it, and closing in on the exact spot earns the rest. Both bonuses are multiplied by A, so a lightning-fast guess on the wrong continent can’t buy back points it didn’t earn.

targets have a size. a guess anywhere inside the target’s radius counts as dead on, and small targets demand tighter guesses than big ones. typical radii: building 500 m · landmark 1.0 km · town 3.0 km · city 12 km · island 8.0 km · region 80 km · natural 10 km · point 5.0 km.

your clock starts on the server when the challenge first loads for you, and stops when your guess reaches the server. scores are always calculated on the server.

example scores
exampleacc.cluesspeedtotal
dead on, no clues, 25 seconds800150501000
40 km off a town, 1 clue, 2 minutes6798528792
40 km off a single building, 1 clue, 2 minutes5987525698
600 km off, all 3 clues, 4 minutes41506421
wrong continent (9,000 km), no clues, 10 seconds6107

rankings

  • the daily board ranks by score. ties go to fewer clues, then less time, then whoever locked in first.
  • the all-time board ranks by total points. ties go to the higher average, then the earlier first game.
  • your percentile (“top 12%”) appears once at least 10 people have played that day. before that you see your exact rank, like “2 of 5 so far”. we never invent player counts or percentiles.
  • your friends board shows you and anyone you’ve swapped friend codes with.

the daily reset

a new challenge goes live at 00:00 utc, every day, everywhere. utc has no daylight saving, so the reset never shifts. a round must be locked in on the utc day it belongs to.

your streak counts consecutive utc days with a locked-in guess. it stays alive until the reset if you played yesterday but haven’t played yet today. miss a full day and it starts again.

past challenges stay in the archive. replays there are practice: they’re scored so you can compare, but never recorded and never change anyone’s official results.

guests and accounts

you can play straight away as a guest. your progress lives in a cookie on this browser. make an account (just a username and password) to keep your streak across devices and appear on the boards under a name of your choosing, or sign in with google. no email, no tracking. the details are on the privacy page.

how we keep the facts honest

  1. source first. challenges are drafted from real reference material, mostly wikipedia and wikidata. the target’s coordinates come from that data, never from an ai model’s memory.
  2. drafting. an ai model may write the hook, the clues and the story, but it has to quote the source word for word for every claim it makes.
  3. automated checks. every quote must appear in the fetched source, the pin is compared with reference coordinates and reverse-geocoded, links must resolve, clues can’t name the answer early, and duplicates are rejected.
  4. a human approves it. an editor confirms the facts, clues, uniqueness and fun before anything is scheduled. a model’s confidence is never treated as verification.
  5. published once. each date is assigned exactly one challenge, and no challenge is ever published twice. the game never asks an ai for a question while you’re playing.

spotted a mistake? every reveal links its sources, so you can check our work.

get in touch

a wrong fact, a bug, an idea for a place, or anything else: email hello@elsewhere.day. we read everything.

credits

map tiles and place search © openstreetmap contributors. facts and coordinates from wikipedia and wikidata (cc by-sa / cc0).