How reputation works
aiternam answers one question: who actually has the best forecasting track record on a topic? Not who is loudest, most followed, or most confident — who has been demonstrably right, over time. This page explains, transparently, how we measure that.
Transparency is the point. A reputation you can't inspect is just another opinion. Everything below is the real method, not marketing.
1. A prediction is a commitment, sealed in time
When you publish a prediction, we stamp it with a trusted timestamp and a cryptographic
fingerprint (content_hash) of exactly what you said. From that moment it is immutable: it
can never be edited or deleted, and the fingerprint proves it. That is what makes a track
record here meaningful — you committed before the outcome was known, and everyone can verify
it.
2. Being right isn't enough — you have to beat the odds
Markets tend to drift upward, so simply predicting "this will go up" over a long enough period is usually right by default. That's not skill.
So we don't score you against "did it happen". We score you against what the asset normally does — its typical direction and its typical swings over your chosen time frame. Your skill is how much you beat that baseline:
- Call the obvious, and you earn little even when you're right.
- Call something the baseline didn't expect — and nail it — and you earn a lot.
- Be confidently wrong, and you lose ground.
When you publish, you also stake a confidence (how sure you are). Higher confidence means a bigger reward when right and a bigger penalty when wrong. Honest calibration pays.
3. Your reputation is a career, not a streak
There are no seasonal resets. Your reputation is built from your whole history of resolved predictions — because telling skill apart from luck takes a real sample. A few lucky hits won't put you on top; a long record of beating the baseline will.
Two things shape the number:
- Recency. Older results gradually fade (with a long, ~2-year half-life), so the leaderboard reflects who is good now, without throwing away the history that proves it.
- Confidence in the score itself. We rank by a conservative estimate of your skill, not your raw average. The fewer predictions you've resolved, the more cautious that estimate — so newcomers appear as provisional until they've built a real track record.
4. Why you can't game it
A few properties fall straight out of the method:
- Volume can't fake skill. Posting more predictions doesn't inflate your rank — your score is an average, not a tally. More predictions simply make your true skill clearer: they prove the good and expose the lucky.
- You can't hide your misses. Every prediction is permanent and counts. There is no delete.
- Popularity stays out of it. Likes, follows and comments are part of the conversation — they never touch your score or the ranking.
5. Proof you don't have to trust us
Two things make the record verifiable, not merely stated:
- Resolution is objective and sourced. Predictions resolve automatically against live market data from a named, fixed source — for crypto, CoinGecko's USD spot price — sealed into each prediction the moment you publish, so the yardstick can never be changed after the fact.
- The seal is anchored on Bitcoin. Each prediction's fingerprint is committed to the Bitcoin blockchain (via OpenTimestamps), so the proof that it existed — unaltered — no longer depends on trusting aiternam. You can verify any prediction yourself, right in your browser.
See also: What makes a prediction.