Every memecoin trader wants the same list: the wallets that are not serial ruggers. It is normally hand-curated, shared as a screenshot, and impossible for anyone to check. This page is how Scienter answers the same question from evidence, what it found, and precisely what the finding does not say.

​
Where the launches come from

Scienter runs a launch census that records a Solana pool when it is created, whether or not it later matters. That ordering is the whole point: a dataset assembled later, from tokens people still talk about, contains only survivors, and every rate computed from it is wrong in the same direction. The census is a sample of launches, not all of them, so every rate below is a rate over that sample.
The census records that a launch happened. It records no outcome and no creator. Both are reconstructed afterwards by the two steps below, and both can fail, which is why “not checked” is a state this data carries rather than a gap it hides.

​
Step one: what became of each launch

Each recorded mint is looked up on a public venue index. Four states, and the distinction between the last two is the one that matters: A run in which nothing answered reports no survival rate, rather than 0%. Zero would read as a finding the run did not earn.

​
When the check happens

A launch is not checked until it is at least 24 hours old, because a token that has not had time to die would be recorded as surviving. There is no ceiling on the other side, and the walk takes the newest eligible launches first, so a caught-up walk measures at close to 24 hours while a backlog is worked through later. In the current corpus the median launch was checked 100 hours after birth, not 24. That matters enough to have been measured rather than assumed. Restricting both sides of the creator comparison to launches checked in a single 24-to-30 hour window reproduces it: the separation holds, non-overlapping, on 2,774 launches. Every row carries observed_at and checked_at, so any figure on this page can be recut by check age without fetching anything again.

​
Step two: who launched it

There is no shortcut. Metadata lookups on these mints return empty creator arrays, because they are bare fungible mints with no creator field to read. What is always true is that somebody paid for the transaction that created the mint. So the creator is the fee payer of the mint’s oldest transaction, found by paging the signature list back to its beginning. A launch whose creation is not reached within the page cap is attributed to nobody. It is never assigned to whichever trader happened to be oldest in the last page fetched, and it is never bucketed under “unknown” — that bucket would become the most prolific creator on Solana and poison every rate here.

​
What was found

A sample of 5,000 consecutive recorded launches, creators resolved and outcomes measured: Grouping those launches by creator: So 225 of the 230 creators with a track record in this window have never had a launch survive, and 153 of the 183 survivors came from a creator who launched exactly once here. An earlier, smaller window of 600 launches reported 42 creators over the bar and zero of them with a survivor. At eight times the sample that became five, and later refreshes read two, then three, then four. On this one it reads five. Those are the same finding measured repeatedly, not several findings: an absolute claim on a small sample softens as the sample grows, and the figure to quote is always the one with the larger denominator behind it. It has now been zero, five, two, three, four and five across six refreshes, which is what a small count does, and this page will keep saying whichever number the rows say. The survival rate moved too — 5.9%, then 2.7%, now 4.7% — and the reason is the denominator rather than the chain: the share of launches never checked fell from 1,420 to 1,078 as the walk caught up on its backlog, so launches that were previously excluded came back in and most of them had resolved. A rate whose denominator is still moving is a rate to quote with its denominator attached, which is why this table states both.

​
The prospective test, which is the strongest evidence here

Everything above is a holdout: the rule was written down, then scored on data that already existed. A fair reader can always ask whether the window was chosen to suit the answer. Since 2026-09-18 the screen has also been running forward. Every time a launch is seen, a call is written to disk — flagged or not — before the outcome exists. The outcome is joined 24 hours later. Nothing about that call can be revised once it is written. Scored 2026-09-21: That is the claim, and it needs no comparison to be useful, because a negative screen’s job is to say what dies.

​
Why no ratio is published for it

The comparison arm — launches whose deployer already had a surviving token — is 456 launches from only 75 deployers, and a single wallet is 18.6% of it. Split by how many of that deployer’s launches were already alive when the call was made: Thirty-one per cent of the rows hold three quarters of the survivors, and that group is eleven wallets. A ratio built on the blend would read as a property of the screen while being substantially a property of one operator, so it is withheld rather than published with a footnote. A number printed and then qualified is still the first number on the page. Counting each deployer once instead of each launch, the comparison is 2.96% against 42.67%, a factor of 14.4. The ordinary unflagged case — one already-live launch, most of the arm — survives at 10.1%, which is what reconciles this page with the 4.16% measured retrospectively on launches checked 24–30h after birth, rather than contradicting it.

​
37.8% of “survivors” are pools nobody trades

Survival on this page means still listed with liquidity. Checked on 2026-09-21 against what those pools were actually doing, that definition turns out to include a large class of tokens that exist and are never traded. Over 108,604 settled launches, 37.8% of everything counted as a survivor had 24-hour volume under 1% of its liquidity. Not one bad actor — 1,368 distinct deployers, and it is concentrated by venue: The extreme case is a single wallet that minted 863 near-identical fluxbeam pools — liquidity clustered around $1.38M against roughly $6,400 of daily volume — and supplied 4.4% of every survivor in the corpus. Excluding untraded pools, the base survival rate over those 108,604 launches falls from 3.63% to 2.25%.
What it does to the screen
It makes the claim weaker and more consistent, which is how you can tell it is a correction and not a tune: The flagged arm barely moves. The comparison arm collapses, because that is where the untraded pools were — and 7.6% sits beside the 4.16% measured retrospectively instead of six times away from it. Thresholds of 0.1%, 1%, 5% and 10% were all computed and the separation holds at every one, so nothing here depends on where the line is drawn. The 1% floor is argued from exit rather than fitted: below it, a position of any size could not be closed without moving the price by more than the whole day’s trading.

​
Do not divide a number from one test by a number from another

The 0.51% above is prospective, at exactly 24 hours. The 4.34% quoted elsewhere on this page is retrospective, at a median check age of 100.5 hours. Survival falls with time, so dividing one by the other produces a figure that describes neither test. The same applies to the 4.16% like-for-like rate, which belongs with its own flagged figure of 0.85%. Each test’s numbers may only be compared with the numbers from that same test: There is no combined figure and no single headline multiple. A claim of “50x” or similar is not supported by anything on this page.

​
Why the site lists more creators than this table counts

This table describes one consecutive window of 5,000 launches, because a survival rate is a claim about a sample and only means something over a defined one. The per-creator pages are built from a different set: every creator with three or more settled launches that the census has ever recorded, currently 4,942 of them. The two numbers differ on purpose. A rate belongs to a sample; a page has to outlive one. Building the pages from this window instead meant that refreshing the data deleted pages that had already been published and linked.

​
What this does not say

This is a record of what happened. It is not a prediction, not a score, and not advice about any future launch.
  • Association, not causation. Repeat launching is associated with total failure in this sample. Nothing here establishes that the same wallet’s next launch will fail.
  • One window. These 5,000 launches are one slice of the census. A separate random sample measured 5% survival rather than 2.5%, so neither figure should be quoted as the rate of anything.
  • Survival is a low bar. live means listed with tradeable liquidity at the moment of the check. It is not profitability, legitimacy, or safety.
  • A creator is not an owner. The fee payer of the creation transaction paid to create the mint. That is not a claim about who controls it now.
  • Delisted is not rugged. Establishing a rug means examining the liquidity-removal transaction. That is a different piece of work and this page does not do it.
  • Surviving is not the same as being worth buying. Measured on this census: entering at a launch’s first reading and exiting an hour later loses about 22% of a $100 clip after costs, and entering at one hour and exiting at twenty-four loses between 14% and 34% depending on the bucket. Every window we can measure is negative.

​
An arithmetic artefact that is not a return

Worth stating because anyone who analyses this data will meet it. Run that same birth-to-one-hour trade without checking whether the position could be sold, and the arithmetic produces a mean of +426%. That figure is not a return, was never available to anyone, and describes five broken rows rather than a market. It is stated here only so it can be recognised and discarded. Why it is false: the ten largest price gains in the sample all occurred against zero liquidity — the largest a nineteen-billion-times multiple against an empty pool, which is a division by a birth price near enough to zero to break the arithmetic. The top five rows account for 100% of the summed price ratio across all 7,215 launches. Requiring only that a $100 clip could be sold into the liquidity present at the exit replaces that artefact with the real figure, −22%. There is no +426% return in this data, on any sample, under any conditions. The same launches produce −22% once the exit is modelled; it is one sample and one trade, measured correctly and incorrectly. Beneath the broken rows the same problem is ordinary rather than exotic: 8.6% of launches have exactly zero liquidity an hour in, only 70% could absorb a $100 exit at all, and among the launches that at least doubled the median liquidity was eight dollars. So a price series on this chain is not a return series. Any study here that does not model the exit will find an edge, and it will not be one.

​
Why a verdict needs three launches

A creator with one dead launch is almost every creator on this chain. Reporting that as a pattern would make the dataset worthless, so below three checked launches the answer is insufficient_evidence and nothing at all is claimed. That threshold is the reason the table above is worth reading: the 42 creators in it each failed repeatedly, across launches somebody actually checked.
Scienter analyses public blockchain records. A finding states what those records show and how confident we are; it is evidence, not proof of anyone’s identity, intent or wrongdoing, and an attribution we cite is the cited source’s. An empty result means not analysed, never safe. Nothing here is legal, tax or investment advice, or a recommendation to transact with any address or asset — see Disclaimers for the full text.