Methodology

Evidence

Every Arvio intelligence output is built from a combination of on-chain data, market data, narrative signals, and where relevant, social signals. Each piece of evidence carries a direction (supporting or opposing) and is surfaced alongside the conclusion so users can inspect the basis for themselves.

Conviction

Conviction is a 0–100 score that reflects how strongly the evidence supports a given view at a point in time. It is not a price target or a guarantee. A high conviction means the evidence base is strong and consistent; a low conviction means the evidence is mixed or thin. Conviction changes as new evidence arrives, and those changes are tracked over time.

Confidence and calibration

Raw conviction is not the same as realized accuracy. Arvio maintains calibration models that learn, per engine, how well past conviction scores predicted actual outcomes. When enough resolved data exists, raw conviction is mapped to a calibrated confidence — an evidence-adjusted probability that the call is right. This keeps the platform honest about how much to trust any single score.

Invalidation

Every thesis carries an invalidation condition: a measurable event that would disprove it. Stating what would prove us wrong is part of the output, not an afterthought. When an invalidation condition is met, the call is resolved accordingly.

Validation and the track record

After the evaluation horizon passes, predictions are resolved against actual market outcomes and marked as a hit, miss, or partial. The full ledger is public on the Validation page. This closed loop — predict, resolve, learn — is how Arvio improves over time.