What Are the Chances This Token Rugs? Paper-Cohort Data

What the fixed paper cohort observed

Correction (September 2, 2026): The headline figures come from a fixed, strategy-selected 53-position Jul 25–Aug 2 cohort mechanically paper-executed at observed live market prices, not funded or live trading. Simulated fills omit some latency, slippage and failure risks. See the fixed-cohort methodology and paper-execution limits.

The figures below come from modeled positions in that fixed cohort. Every decision and paper outcome was logged, but no funded order was placed.

Every token underlying a modeled position in the fixed, strategy-selected cohort had already passed the strategy's full safety gate: LP burned at graduation, mint and freeze authority revoked, no hostile Token-2022 extensions, holder concentration within limits, creator wallet not blacklisted. These were selected survivors of screening, not random launches.

• Within the fixed, strategy-selected paper cohort, 8 of 53 modeled positions (15.1%) were rug-tagged despite passing the strategy's checks.
• Within the fixed, strategy-selected paper cohort, the paper profit share was 57% — 30 of 53 modeled positions closed with simulated profit.
• Within the fixed, strategy-selected paper cohort, median time to rug tag was ~89 minutes from modeled entry near graduation.
• Within the fixed, strategy-selected paper cohort, 90th-percentile time to rug was ~10.6 hours.
• Within the fixed, strategy-selected paper cohort, rug-tagged modeled positions averaged a −72% simulated loss, with a 12.7-minute median modeled entry-to-close interval and a 23-minute average.

What the separate creator subgroup observed

Creator history was also examined in strategy-selected paper-executed entry data. The 69-position first-launch subgroup below is separate from the fixed 53-position cohort and should not be read as a funded result, population rate or causal probability:

First-launch creators are unproven because their tokens have no history. Separately from the paper cohort, the engine has rejected 1,837 serial-creator graduations, including 336 from wallets with 50+ prior launches. The current blacklist of confirmed ruggers — 53 wallets — has blocked 252 candidate entries. Those production screening counts should not be combined with the fixed cohort to assign a probability to the blocked entries.

Full creator-check methodology: How to Check a Solana Token's Creator Wallet.

Why unscreened launches are a separate question

The fixed paper cohort included only tokens that cleared the strategy's safety checks, so it cannot establish the prevalence of rugs among unscreened launches. Academic research on broader DEX launches — including Mazorra et al.'s "Do Not Rug Me" — uses different definitions, chains and samples. Those external findings provide context but cannot be combined with this cohort into one probability.

Pump.fun's graduation filter itself acts as a partial screen: only ~1-2% of all tokens launched on the bonding curve ever graduate to Raydium. Graduation requires buying pressure to move the bonding curve, which filters out the cheapest, fastest ghost launches. The paper cohort included only tokens that had cleared this hurdle.

Risk factors that require separate evidence

No single factor is deterministic. The following checks can identify mechanisms or concentration that deserve scrutiny, but this fixed cohort does not estimate how much each factor changes a token's probability:

High holder concentration

When the top 10 wallets (excluding pools) hold most of supply, a small group can move the market sharply. The separate holder-concentration study states its own cohort, outcome definition and limits; it should not be merged with these 53 paper positions.

Bundled launch

High bundler counts — wallets funded from one source buying in the launch block — were our 4th most common rejection flag (160 rejections in one week of telemetry). Bundled supply behaves exactly like a concentrated wallet when it's time to exit; the number of addresses is theater.

Token-2022 extensions

Transfer hooks and permanent delegates were logged in 58 and 45 rejections respectively. Active configurations can restrict transfers or give another address control, but those rejection counts do not measure a rug probability or prove that every extension is malicious.

Copycat tickers and narrative-chasing launches

A review of the eight rug-tagged paper positions found more copycat tokens than expected — launches riding another coin's ticker within hours of it trending. This small cohort observation supports extra scrutiny but does not establish a population-wide rate.

What eight rug-tagged positions do and do not mean for sizing

The eight rug-tagged outcomes demonstrate that severe modeled losses remained possible after this strategy's screening. They do not provide odds for the next token. Real orders can receive materially different fills, so sizing should be based on a trader's ability to absorb a total loss rather than treating this selected cohort's share as a forecast.

Within the same paper cohort, mechanical take-profit exits averaged about +11%. That comparison is descriptive and unweighted, not evidence of funded strategy performance. Full exit framework: When to Sell a Solana Meme Coin.

How MemeAssist grades rug risk

The Rug Risk score in a MemeAssist report synthesizes the above factors into a single headline number: holder concentration, mint and freeze authority status, LP burn status, bundler count, creator history, wash-trade ratios and liquidity depth. Paste any Solana mint address in the analyzer to compare its profile with the eight rug-tagged positions in the fixed paper cohort, while recognizing that the cohort does not provide a token-specific probability.