Navigation — a beginner's tutorial

1 · What is this game?

You are somewhere on Earth and you want to reach a target territory — a country, a region, sometimes a single point. You get there by flying a sequence of legs: each turn you choose a compass heading, and the game flies you that way along a great-circle arc for a fixed distance. The catch is that the legs shrink every turn, and your position is hidden from you — you only get a masked report with clues.

You're racing an opponent to the same target. The first to arrive wins. So the game is part navigation puzzle (where am I, and how do I steer there?) and part race (can I get there before they do, and if we tie, who wins?).

2 · The board & the goal

The board is the real WGS84 Earth. Distances are in metres (we'll often use megametres, Mm = 1000 km), bearings are degrees clockwise from north (0° = north, 90° = east, 180° = south, 270° = west).

• The target territory is announced at the start of the round (e.g. “metropolitan Denmark”, or “north of 89°N”). To land is to reach it — usually you must land on land.
• The beacon is a single point that lies inside the target. Your report tells you roughly how far you are from it. It never tells you the direction directly — you have to work that out.
• Your start is wherever you finished the previous round. (The very first round often starts you at a known place — though not universally.)

You don't need to land on the beacon — just anywhere in the target territory. The beacon matters mostly for the tie-breaks (see §8).

3 · A turn, step by step

Every turn has the same shape:

1You receive a location report. A masked snapshot of where you are: distance clues to the beacon, which country/sea you're over, your timezone, and what the opponent is doing.

2Optionally, you use the radio. Spend some of your limited “tape” to record nearby radio stations and triangulate your true position. Optional and costly — most turns you won't need it.

3You submit a heading. One number in degrees. The game adds a random wobble to it (see §4) and flies you one leg along that bearing.

Then the next report arrives and it repeats. Your thinking time runs on a chess-style (Fischer) clock: you have a bank of time, each committed move adds a fixed increment, and whatever you don't use carries over — so moving quickly on easy turns banks time for a hard one later. You can post a safe “fallback” heading immediately and keep refining it until the deadline — see the interface tour.

4 · The three things that move you

Heading — the one number you control

You pick a bearing in degrees. That's your whole move. Everything else is fixed by the rules or by chance.

Leg length — fixed, and shrinking fast

Each leg is the previous one divided by the golden ratio φ ≈ 1.618. The first leg is a whopping 12.36 Mm (a third of the way around the planet); after that they fall off quickly:

Two consequences worth internalising early:

• Leg 1 almost always overshoots. The target is usually much closer than 12.36 Mm, so your first leg flies right past it. Turn 1 is a setup move, not a landing attempt — don't fight this, use it.
• You can't fine-tune with a big leg. Landing on a small territory needs a small leg, because a big leg turns a tiny heading error into a huge miss. Precision arrives as the legs shrink.

Deviation — the random wobble

Here's the twist that makes the game hard. The heading you submit isn't flown exactly. The game adds a random deviation drawn from a Cauchy distribution (scale 7.5°). Concretely:

• Half the time the wobble is under 7.5°. That's already a lot on a big leg.
• It is heavy-tailed: occasionally the wobble is enormous — 45°, 90°, more. These aren't bugs; they're the distribution. Plan for them.

On leg 1 (12.36 Mm), even a modest wobble throws you thousands of kilometres. And it can be much worse than modest: about 1 draw in 100 spins the heading more than a full turn around the compass, landing your leg in a completely unrelated direction. This is the reason you can't just “aim at the target and go”.

But the heavy tail is a curse and a blessing. Curse: no single plan survives a wild kick. Blessing: your opponent flies the same wobble, so it never favours them; a big kick can fling you onto useful, identifiable land you'd never have reached by aiming; and — the key one — once you know the seed you can cancel the wobble entirely. That last point is the single most important idea in the game, and it's next.

5 · Reading your location report

Each turn's report is a masked view of your position. You never get raw coordinates. The fields you'll use most:

A useful habit: read inferred_bounds and the country/sea first (“how far, roughly where”), then the opponent line (“am I winning the race?”), then decide whether you need the radio.

When the layers disagree, know which to trust: esri has the most accurate coastlines and international borders, so it's your best judge of landfall (on land? which country?). natural_earth takes a Swedish political view — it recognises, say, Kosovan sovereignty — and is often closest to how the target territory is defined, so use it to decide whether a spot counts as the target.

6 · The seed: your single biggest edge

The random deviations aren't truly random — they come from a pseudo-random generator seeded once per round. If you can figure out that seed, you can reproduce every deviation for the rest of the round in advance. And if you know the wobble ahead of time, you can cancel it: submit heading = (the bearing you actually want) − (the known deviation), and you fly exactly where you intended. Perfect steering.

How the seed works, and why it's recoverable:

• The seed is a timestamp — the round's creation time, formatted as an Oslo-local date string — jittered by a random amount. The game tells you the mean (seed_mean) and spread (seed_std) of that jitter.
• So the seed is one of only a few hundred / few thousand candidate timestamps around the mean. You can enumerate them all, and for each one reproduce the deviation stream it would produce.
• Then you filter: each candidate predicts a different path, and therefore a different sequence of free clues (on land or sea? which country? which timezone?). Cross off every candidate whose prediction doesn't match what your reports actually said. After a few turns, usually one candidate survives.

This is the game's core skill. The recommended order of attack:

1Free filtering first. Use the land/sea, country and timezone from each report to eliminate candidates. Costs nothing, works everywhere, and often pins the seed by turn 3–4 on its own. The Seed recovery tool does this for you.

2Distance bands next. The ln_dist ring each turn also filters candidates — even over open ocean where there's no country to go on.

3Radio only to break ties. If two or three candidates survive with identical free clues, a single radio fix can tell them apart. That's a last resort, not a first move.

7 · The radio — a scalpel, not a hammer

The radio lets you record nearby airport “stations” and triangulate your true position. Each station broadcasts a fixed signal that fades with distance; you record a bit-stream of which stations you hear, and infer where you must be. It costs tape, a small budget issued each round, so you can't use it every turn.

Two ideas make it worth much more:

• Range is free and decides everything. You tune a “range” (in metres) that sets which shell of stations you can hear. Too small and one nearby station drowns out the rest (you learn a distance, not a position); too large and everything sounds the same. Tune it to the spacing of the airports around you so several stations sit at different bearings and audible strengths. Most wasted tape is mistuned range, not too few recordings.
• A free dry run scouts first. You can record zero tape to see which stations are nearby before spending anything. Do this before any real call.

When should you spend tape? When it will change a decision — typically to break a tie between surviving seed candidates while you're over dense airport country, or in the endgame to be sure which side of a coastline you'll land on. Over open ocean with huge legs, a precise fix barely changes your heading (the wobble dominates), so save your tape. Think of tape as something you hoard for the one or two turns where geometry is rich and the stakes are high.

The radio is getting a directional-antenna upgrade (steerable beams via “harmonics”) — see the Wiki for the full model when you're ready. As a beginner, an omnidirectional call tuned to the local airport spacing is all you need.

8 · Choosing when and where to land

There are exactly three ways a round is decided. Knowing them changes how you play from turn 1:

1First to land wins outright. No tie, no argument. So the biggest single lever is landing one turn earlier than your opponent. Always ask: is a landing available this turn? Measure its real margin before dismissing it as risky.

2Same-turn tie → the player farther from the beacon wins. If you and the opponent both land on the same turn, whoever is farther from the beacon takes it. So if you can't land first and a tie looks likely, aim for the point in the territory farthest from the beacon — while keeping a safe margin inside the border.

3Everyone strands → closest to the beacon wins. If nobody can reach the target, the consolation prize goes to whoever ends up closest to the beacon. So if you conclude you can't land, switch goals and close in on the beacon instead.

The opponents field is your live scoreboard for conditions 2 and 3: closer_to_beacon tells you who'd win a same-turn tie right now, and a stranded status means you win simply by landing at all. But treat closer_to_beacon as a snapshot, not a settled result — it flips easily. If your next leg overshoots the beacon and comes down the far side, you can go from farther to closer (or the reverse) in one turn, so re-check it at the turn you actually land.

9 · Your first round, narrated

Suppose the target is a mid-sized country a few megametres away. Here's the shape of a competent round — not the only way, but a reliable template.

Turn 1 (setup). Leg 1 is 12.36 Mm and will overshoot no matter what, so don't aim to land. Instead choose a heading whose likely landing spot (after the wobble) is over identifiable land with airports — this maximises the free clues that pin the seed, and keeps the target reachable. Post a fallback heading right away, then refine.

Turn 2 (gather & narrow). The report's country + timezone eliminate most seed candidates. Feed your observations into the Seed recovery tool. Aim your next leg to keep the target reachable and, again, to land somewhere identifiable. If you're over dense airport country and only a couple of candidates remain, this is where a radio tie-break can pay off.

Turn 3 (pin the seed). With two turns of clues, usually a single seed survives. The moment it does, your position is exact and all future wobbles are known. Now switch from “gathering” to “finishing”: compute the earliest turn on which a small-enough leg can put you in the target.

Turn 4 (land). Cancel the known deviation so you fly exactly the bearing you want, and land — at the earliest possible turn to win outright, or (if a tie looms and you're the closer player) at the in-territory point farthest from the beacon. Keep a margin inside the border so a small position error doesn't drop you in the sea or the wrong country.

10 · The interface, guided

The player page (Control room) is organised around the current turn:

• Turn cards, newest first, each centred on that turn's location report. Unread events sit at the top until you dismiss them.
• The countdown. When a turn has a deadline, the hub shows a live “submit by … · in …” timer that changes colour as the deadline nears, alongside the turn's increment. Remember it's a Fischer clock — time you don't spend carries into the next turn.
• The submit-flight form with a Provisional toggle. Check it to set a valid_from time: your move stays editable until then, so you can post a safe fallback heading immediately and overwrite it in place as your analysis improves. This is the intended way to use the whole time budget without ever risking a missed turn.
• The radio form with a map (great-circle range rings at 1×/2×/3×) and a directional-gain plot for when you use harmonics.

Two client-side tools live in the navbar (they run entirely in your browser):

• Seed recovery — sweeps candidate seeds and scores them against your observed deviations. Your day-to-day seed-pinning workhorse.
• Map — a scratch map with geodesic range circles that snap to the golden-ratio leg lengths, bearing ticks for reading off headings, and GeoJSON export. Think of it as a straightedge and compass: the leg circles are your compass arcs and the bearing ticks your straightedge, so you can construct an entire trajectory — or a landing geometry — by hand, no code required.

11 · Common beginner mistakes

• Treating turn 1 as a landing attempt. It overshoots. Use it to set up clues and geometry.
• Aiming straight at the target every turn. Sometimes correct, but it can walk you into a trap where a big leg flings you back out, or it wastes the chance to land somewhere that pins the seed.
• Ignoring the seed. Without it you're at the mercy of the wobble. Recover it early; it's the difference between guessing and steering.
• Trusting a far ln_dist. Far readings are almost useless for locating the beacon. Wait for a close one, or triangulate from several.
• Spending tape out of habit. Radio only when it changes a decision. Do a free dry run first, and get the range right before the sextets.
• Landing dead-centre into a tie you'd lose. Watch closer_to_beacon; if you're the closer player and a tie is coming, land toward the far edge.

12 · Where to go next

• Wiki / reference — exact numbers, the full field reference, the radio's signal model and directional antenna, win-condition details, and a glossary.
• Seed recovery tool — start feeding it your deviation observations from turn 1.
• Map tool — plot your legs and plan a landing.
• Control room — play.

This tutorial will keep growing as we learn more about what helps new players. If a concept here felt thin or confusing, that's worth flagging.