Memory, not fortune telling.

Every coin is photographed at exactly these ages, so any two coins can be compared fairly:

Market cap, liquidity, volume, buys and sells, unique buyers, holder count and distribution, bundler and insider share, curve progress, plus the coin's name, image and narrative. Anything a source cannot confirm is shown as UNKNOWN — never filled in with a guess.

SORA keeps watching each coin and records what actually happened: peak market cap, multiples from each checkpoint, drawdown, whether it graduated. A coin is only counted as a runner once the real path says so.

A similarity score is a distance measurement, not a probability. 80/100 means this coin's numbers at this age sit close to where past runners' numbers sat at the same age. Coins that look identical still fail constantly — that is why SORA also shows how much a coin resembles past failures, and always lists closest matches that went nowhere.

No snapshot may contain data from after its own timestamp. Models are tested against coins newer than the ones they learned from. Rebuilt histories are labelled as rebuilt, never as something SORA watched live. And every score states how many coins it came from.