Every scanner sells liquidity depth as a safety signal. Across 372,209 pump.fun launches, it isn't one. Split the coins that actually traded into five bands of early bonding liquidity — a 12× span — and the failure rate barely moves: 88–94% rug or die at every level. Deeper liquidity doesn't make a token safer. It makes it worth rugging.
Split the 69,238 coins that actually traded into five bands by early bonding liquidity, read at 5 seconds — a 12× span, from under 0.85 SOL to over 10.6 SOL. The failure rate: 93.6%, 88.0%, 89.3%, 92.7%, 89.7%. It barely moves, and it isn't even monotonic.
Liquidity does buy something — a 12× better shot at graduating, 0.17% to 2.04%. It buys no protection at all. Deep liquidity doesn't make a token safe. It makes it worth rugging. That is the signal every other scanner is selling you.
Across 372,209 pump.fun launches — 72% of every curve we captured — each tracked to a recorded outcome. 96.8% rugged or died. 65% never even cleared a 2-SOL peak; every one of those died.
Of all 372,209 launches: 0.13% ever graduated. 89% never doubled.
Two populations, both stated, never blended. The 372,209-launch census carries a recorded outcome for every coin. The bars above are the 1,944 coins with full trade-level labels — the only subset with a realized return — and all four bars, including liquidity, are computed over that same 1,944. "Sellable at a profit" = peak REAL liquidity ≥ 10 SOL: enough depth for a modest exit. Every n is disclosed. This is the layer risk-scanners never measure.
Outcomes labeled in a single pass on 2026-06-28, covering launches through 2026-06-27. Capture has since run to 2026-07-09; those launches are not yet labeled. Live scans are real-time.
Every scanner grades a token in the present tense. None of them carry the outcome. Forward-labeled history is the asset nobody else publishes.
A present-tense flag. Useful, binary, and already free from several funded tools.
The recorded track record — counts, medians, timings — pulled from outcomes already labeled in the capture.
Compared against 63.7% — the rug rate across every launch we can attribute to a creator, measured the same way. Not the 28% figure for all 372,209 launches: 65% of those never clear a 2-SOL peak and rug at 0.0% — they cannot rug, they die. And we attribute rugs more easily than quiet deaths: pulling liquidity leaves a signed transaction, a token that fades leaves its dev anonymous. Reweighting for liquidity closes only 0.4 of a 6.5-point gap, so creator rug rates here run roughly 6 points hot. We publish the bias rather than quietly absorb it.
GROUNDTRUTH replays what actually happened to a pump.fun coin — not the chart's best case, but what you could really have sold into. Paste a coin for its autopsy, or a wallet for its rap sheet.
Drop a token contract to see who's holding it right now: holder concentration, snipers still in, how many known ruggers and flagged fast-exit wallets are already in front of you — plus the launcher's rug history. Or drop a wallet address to pull its rap sheet.
Realized exit — what you'd actually net selling into real liquidity, not the chart price.
Peak real liquidity — the SOL truly available to sell into at the coin's deepest.
Verdict — the recorded outcome: rugged, bled out, graduated, or survived.
// paste a coin (CA) or wallet above to run a live scan
The live feed and the wallet/creator checker are opening to a first batch. Get in before it's gated.
Or follow the data drops on @groundtruth_now — the findings post there first.