Why being right isn't enough

This is the one idea that makes aiternam different from every "I called it" screenshot on the internet — and the one that surprises people most. So take it slowly.

Being right is not the same as forecasting well. Markets drift upward over time. Predict "Bitcoin will be higher a year from now" and you'll be right more often than not — not because you have insight, but because that is simply what the asset tends to do. A track record built on calls like that measures nothing.

So aiternam does not score you against "did it happen?". It scores you against what the asset normally does — and rewards only the part of your call that beats that.

Meet the skill-free robot

Imagine a robot that knows nothing about the future and has no opinions. All it knows is the asset's own history: which way it tends to drift, and how much it typically swings. Ask it "will Bitcoin be up in a year?" and it answers with a probability drawn purely from that history — call it the drift-aware baseline.

Your score is, quite literally: did you out-forecast that robot?

The robot already knows crypto tends to drift up. So an "up, long horizon" call barely impresses it — it was already fairly sure. You beat it by only a little, even when you're right.
Call something the robot thought unlikely — a sharp move, a reversal, the direction against the drift — and nail it, and you beat it by a lot.
Be confidently wrong, and you do worse than the robot, and you lose ground.

The robot is the baseline. Everything on aiternam is measured as distance from it.

Where the baseline comes from — and when it's fixed

The baseline is built from two numbers about your asset, estimated from its price history at the moment you publish, using no information from after that moment:

its drift — the return the asset tends to produce over your time frame;
its volatility — how much it typically swings over that same time frame.

From those, the robot's probability that your side comes true is fixed.

The difficulty of your call is sealed at creation, inside the same fingerprint that seals the forecast itself (see Sealed in time). Neither you nor we can renegotiate it afterward — nobody gets to decide, once the outcome is in, that your call was "obvious" or "a long shot." The robot's odds were locked before anyone knew the answer.

The obvious call earns almost nothing

Say you call BTC up over one year at 80% confidence. At publish time, the asset's own drift and volatility put the robot's probability of "up" at about 65% — a rising asset, so the robot already leans your way.

You're right: a year later BTC is up. But you were only a little more confident than a robot that already expected this. Your skill score for that call is small and positive — real, but modest. You beat the baseline by the gap between your 80% and its 65%, no more.

Now take a braver call. Suppose the robot only gave your side 35% — you were forecasting against the drift, something it thought unlikely — and you nailed it at 80% confidence. Same outcome for you (correct), but you beat a robot that was leaning the other way. That call is worth far more.

The rule of thumb. Skill is the surprise in a correct call. The more the baseline doubted you, the more a correct call is worth — and the more a confident miss costs.

Being right with a negative score is possible — and correct

Here is the part that trips people up. You can be right and still lose points.

If the robot gave your side 65% and you published at only 55% confidence, you were less sure than the skill-free baseline. Even when the asset moves your way, you forecasted worse than the robot did — you left conviction on the table on a call the baseline already favored. A small negative score is the honest verdict, and the system is behaving exactly as intended.

This is not a bug to complain about. It is the whole point: the number rewards beating the baseline, not being on the right side of a coin the market was already going to flip your way.

Why the permabull never wins

The clearest test of the method is the person who predicts "up" on everything, always, at maximum confidence. In a rising market they'll rack up a wall of green correct marks.

They still don't climb. Every one of those easy "up" calls is measured against a robot that also expected "up" — so each beats the baseline by almost nothing, and their average skill drifts toward zero. And when the market turns, their confident misses are punished hard. Over a career, a permabull's score converges to no demonstrated skill, which is the truth about them.

That is the promise of this page: on aiternam, being right isn't enough. You have to be right about something the market didn't already know — and the more you do that, the higher you rise.

Next: Confidence is a stake — what the confidence number does to the size of that reward or penalty. Then How a prediction resolves — what actually happens at your time frame. For the map of the whole system, see How reputation works.