How a prediction resolves
Every prediction has a deadline: the time frame you chose when you published. When that moment arrives, aiternam resolves it — decides whether you were right, and by how much. This page is exactly what happens then, including the one outcome people are surprised by: the void.
Resolution is automatic and objective
There is no human in the loop, and no judgment call. When your time frame is reached, the scoring engine:
- reads the outcome price from the same market-data source that was sealed into your prediction at publish time — for crypto, CoinGecko's USD spot price;
- compares it to the entry price that was captured and sealed the moment you published;
- decides correct, wrong, or void, and computes your skill score.
Because the source and the entry price were sealed at creation, nobody — including us — can pick a friendlier price after the fact. The yardstick was chosen before your call, and it never changes. (More on the sealed source in What you can predict on.)
The noise dead-band — why a "correct" tiny move doesn't score
Here is the outcome that generates the most support questions: "The price went my way — why didn't I score?"
Because a move can be too small to mean anything. If you call a volatile coin "up" and it closes 0.3% higher, you didn't forecast anything — that's the asset breathing. Crediting it would reward noise and let people farm points on coin-flips.
So around the entry price there is a dead-band: a zone of "this didn't really move." A bare directional call that lands inside the dead-band is a void — it counts for nothing, neither right nor wrong, and never touches your reputation.
The dead-band is not a flat percentage. It is scaled to your asset over your time frame: roughly half of the asset's typical swing for that horizon (its sealed volatility). A placid asset has a narrow band; a wild one has a wide band. So "a move worth noticing" means the same thing — beyond the noise — whether you called a stablecoin-quiet asset or a volatile one.
Want the dead-band. It is on your side. It protects you from being marked wrong by a meaningless wiggle against you, just as it declines to credit a meaningless wiggle for you. A void is the system admitting the market didn't actually answer your question.
Voids don't apply to target-price calls
If you name a precise target price, the dead-band doesn't apply. A target call asks a sharp, unambiguous question — did it reach this level? — and the answer is always yes or no. There is no "too small to tell": either the asset touched your target or it didn't. That is also why a target call can never void, which changes the risk in a way worth understanding on its own page: Directional vs target-price.
When there's no data — the unresolvable void
Occasionally an asset can't be priced at the deadline — it was delisted, or the data feed has no value for it. The engine retries for a while; if the horizon has passed and the asset is still unpriceable, the prediction is voided as unresolvable rather than left hanging forever. Like any void, it counts for nothing. A prediction with no valid entry price to begin with can never be scored and voids immediately.
The resolution layer is appended, never edited
When a prediction resolves, the outcome — the closing price, whether you were right, your score — is added to it as a separate resolution layer. Your original forecast is not touched. Not one field of what you sealed is altered; the result is written alongside it.
This matters for trust. The immutable thing is the prediction — your commitment, made before the outcome was known. The resolution is derived from it afterward and recorded openly. So a resolved prediction shows both halves honestly: what you said, sealed and unchanged, and what happened, appended and dated.
Next: Confidence is a stake — how the confidence you set sizes the score you get. Or step back to How reputation works.