Holder counts are read from the chain. A plain number is exact; a ≥ number is a floor, where the token's history was too deep to walk and concentration is left blank rather than guessed. Tap any row to see why it scored that way; tap a contract or dev address to copy it.
Rug score is out of 10 and marked prior: a heuristic, not a measured probability — 7+ is high, 4-6.9 elevated. Each signal has a points budget and spends part of it, so the parts add up to the total; a token measured on 4 signals and one measured on 9 are both scored out of 10. — no rug has been labelled yet, so nothing is fitted. Hover or long-press the score to see which signals drove it. It is replaced by a fitted probability once labels accumulate; rugs are far more common than 2× pumps, so that arrives well before the pump model.