About this site

orangekrakenlab publishes forecasts about the physical supply of critical materials — the routes they travel, the places they can be choked off, and the technologies that will need more of them — and then keeps score in public. Every number on this site is either a forecast that was locked before the outcome was known, or the grade that outcome later earned. This page explains the few words you need to read the rest.

A frozen call

A frozen call is a specific, measurable forecast written down before the outcome exists: what is claimed, the probability we give it, the date it will be judged, and the public source that will settle it. Once frozen, a call is never edited, re-dated or quietly dropped. Each day’s freezes are hashed, committed to a public git history and anchored in a Bitcoin block, so an outsider can check that a call really was made when we say it was — see Verify.

Graded against a stated null

Every call is scored against a null — the dumb answer you would get without any model. For weather-like questions that is climatology (how often the thing happens in the long run); for slow-moving supply questions it is persistence (nothing changes). The null is chosen and written down at freeze time, never afterwards. Probabilistic calls are scored with the Brier score (lower is better); a lane only claims skill when its Brier beats its null on the calls that have actually matured. Where a professional forecaster publishes its own probability for the same event, that number is captured beside ours and shown as the official rival.

Matured, and earned in

A call is matured when its judgment date has passed and the outcome can be read from the named source. Matured calls are graded mechanically; if the observation needed to grade one was never captured, it is closed as unobservable, counted apart, and never quietly turned into a hit or a miss.

A model, a lane or a signal earns in only when it clears a bar that was fixed in advance — typically a minimum number of graded calls and a better Brier than its null on those calls. Until then the site shows the null’s number, labelled as such. Nothing is promoted because it looks promising; things are promoted because their graded record says so.

The OKLM

The OKLM is a small learner trained only on this site’s own graded record — the frozen calls and the outcomes they earned. It is scored exactly like everything else: each lane reports how many graded rows it has, its forward Brier against the null, and the rung it has reached; a lane that has not beaten its null says so. Its ledger is on the Ledger page.

The corridor focus

The home page is organised around distribution corridors: the routes critical materials actually travel (straits, canals, ports, rail, roads and border crossings), the chokepoints where a single disruption would bite hardest, the alternatives available if a corridor degrades, and three separate facts about each alternative, which this site keeps apart because conflating them overstates a detour: its shared legs (legs the two routes genuinely travel in common), its shared places (ports and passages both routes call at, even by different legs), and whether it actually escapes the specific chokepoint a question is about. Until 2026-09-06 these were one list, and a detour with four of the primary’s six legs was described as sharing all six. The Materials page models who produces and controls each material; Robotics and Quantum read future demand early; Sensors watch the physical world the corridors run through and grade in days, which is where calibration is earned fastest.

Why this exists — in plain English, and for the kids

* In plain English

This site makes predictions about whether the world’s supply of important materials is about to get harder — and it makes them in a way that can be checked. Each prediction is written down and locked before anyone knows the answer, then scored later against what really happened, next to the score a “nothing changes” guess would have got.

Wrong calls stay on the record with their grade. Nothing is deleted, nothing is edited after the fact, and nothing is called skill until it has beaten the dumb baseline enough times to matter.

If it works: A public track record you can trust when it says trouble is coming — because it proved, in the open, that it called earlier trouble before the outcome was known.
For the kids

Imagine guessing whether it will rain tomorrow, writing your guess in permanent marker, and then checking the next day. That’s what this site does — except the guesses are about the special metals inside phones, robots and computers, and the ships and roads that carry them.

The important rule: no erasing. If a guess is wrong, it stays on the wall with a big “wrong” next to it. That’s how you can tell whether the guesser is actually getting better.

If it works: A lookout you can really trust, because it showed its guesses BEFORE the answers came in — not after.

Disclaimer

Nothing on this site is investment, safety or operational advice. Corridor and chokepoint geography is descriptive — it says where things are and what depends on them, not what anyone should do. Every item names the public source it is graded against; where a source is missing, the item says so. Forecasts are probabilities, and a probability is not a promise.

*In plain English: Read the grades, not the headlines. A calibrated forecaster is right about as often as it says it will be — which means it is also wrong exactly as often as it says it will be.
For the kids: Grown-ups: this is a scoreboard, not a rulebook. Kids: it's a guessing game with receipts — nobody should make big decisions just because a guess on this site said so.