Dead Coin Statistics: Why Every Estimate Disagrees

Ask how many cryptocurrencies have died and you get answers three orders of magnitude apart. Coinopsy has catalogued roughly 2,400. 99Bitcoins lists 48,351. CoinGecko counts 13.4 million.

None of them is lying. They're answering different questions, and almost every "half of all crypto is dead" headline you've read is missing the sentence that would make its number mean something.

The number everyone quotes

The current headline belongs to CoinGecko: 53.2% of all cryptocurrencies on GeckoTerminal have failed. Its per-year counts run 2,584 failures in 2021, 213,075 in 2022, 245,049 in 2023, 1,382,010 in 2024, and 11,564,909 in 2025 — 13,407,627 in total, for data between 1 July 2021 and 31 December 2025.

The same page reports that listings grew from 428,383 projects in 2021 to "nearly 20.2 million" by 2025.

Those two figures don't fit together. 13,407,627 out of 20.2 million is 66.4%, not 53.2%. CoinDesk prints both halves of the contradiction in a single sentence: "Of the nearly 20.2 million tokens that entered the market during that period, 53.2% are no longer actively traded."

We can't tell you which number is wrong. Either the denominator isn't 20.2 million or the percentage isn't 53.2% — the published figures don't say which, and guessing would be exactly the sin this page is about. What we can say is that the most-cited dead-coin statistic in circulation doesn't survive a calculator, and that nobody quoting it has checked.

An earlier cut of the same study did reconcile. With data through March 2025, Cryptopolitan reported 3.7 million failures out of roughly 7 million — 52.7%, and 3.7 ÷ 7 lands where it should.

There's a second thing the headline omits. CoinGecko counted only tokens with at least one trade, and among pump.fun tokens, only those that "graduated" — cleared a market-cap threshold and moved to a real DEX. Graduation is rare: verified rates ran between 0.37% and 1.78% in April 2025, and just over 1% in January 2026. So roughly 98–99% of launchpad tokens never entered the sample at all. That makes 53.2% a generous reading — it's mortality among tokens that already cleared a liquidity bar, not mortality among tokens. (That inference is ours, drawn from the two rate sources; CoinGecko doesn't frame it that way.)

"Dead" is a snapshot with a timer

The clearest demonstration that definitions drive everything comes from one dataset producing two famous numbers.

In The Rise and Fall of Cryptocurrencies (WEIS 2018), Feder, Gandal, Hamrick, Moore and Vasek tracked 1,082 coins over nearly five years, defining abandonment as "a drop in average trading volume to below 1% of a prior peak value." They found 44% were abandoned at least temporarily. They also found 71% of those came back. Which leaves 18% permanently dead.

44% or 18%, same coins, same authors, same paper. The only variable is whether you stop counting on the day the volume dies or wait to see if it revives. Cite one without the other and you've doubled or halved the story.

Their 2021 follow-up added tokens to the picture and found they behave differently: 44% of coins abandoned against 7% of tokens; 18% of coins permanently dead against 5% of tokens.

Meanwhile the trackers use a rule that appears nowhere in the academic literature. CoinMarketCap Academy describes the convention: platforms "consider a cryptocurrency dead or abandoned if it has had a trading volume of less than $1,000 within three months." Coin Kickoff uses the same test.

So: under $1,000 in three months, below 1% of peak volume, or no active trade since. Three tests, three populations, three answers. None of them is wrong.

Why every estimate disagrees

Our number is the smallest on this table by orders of magnitude, and that's deliberate. It answers "which coins died, and why" — not "how many." Every entry names a cause and carries a note; where a stable reference exists, we link it. You can read the whole database and check us entry by entry, which is not true of any count above it.

Two of those rows deserve a warning label. 99Bitcoins publishes 48,351 with no stated methodology and no source, under a headline that rounds it to "40k+" and a title claiming a 2026 update with no date attached. And Coin Kickoff's Coinopsy analysis states plainly that its data was compiled in August 2022 — on a page dated 8 January 2026. Three-and-a-half-year-old numbers, presented as current, in a market where 86% of all deaths happened after the data was collected.

The ICO cohort

The 2017–18 ICO wave is where dead-coin statistics were born, and it's a good lesson in how a number goes feral.

The famous one: 44.2% of ICO startups were still alive 120 days after their raise, from Benedetti and Kostovetsky's working paper on more than 4,000 ICOs that raised $12 billion between January 2017 and March 2018. Survival was measured by tweeting. Projects with zero Twitter activity five months out showed a 100% failure rate. The average raise was $11.5M, the median $3.8M, and only 48% of the sample reported raising anything above zero.

Worth knowing: when that paper was published in the Journal of Corporate Finance in 2021, the sample was 2,390 completed ICOs and the abstract led with returns — average 179% from ICO to first exchange open — not with mortality. The mortality statistic that made the headlines lives in the preprint.

Then there's the Satis Group finding: "approximately 78% of ICOs were Identified Scams, ~4% Failed, ~3% had Gone Dead, and ~15% went on to trade on an exchange." That became "80% of ICOs were scams" and never left.

The critique lands hard. Crypto Briefing pointed out that "only ~$1.3B (~11%) of ICO funding went to Identified Scams" — and CryptoSlate, reporting the study favourably, noted the same thing: fraudulent ICOs took less than a tenth of the money. 78% of the projects, 11% of the capital. The industry ran with the first figure for years. (Use that article only for its arithmetic. Its project-level judgments were badly wrong — it called AriseBank "an earnest effort by naive founders with no stolen money," which the courts would shortly demolish.)

And the ICO totals themselves rest on self-reported numbers. In January 2018 the SEC halted an ICO "claiming to have raised approximately $600 million from investors since November 2017" — the emphasis is the SEC's own hedge. AriseBank's Jared Rice Sr. later pleaded guilty and was sentenced to five years, ordered to pay $4,258,073 in restitution. The real raise was roughly 1/140th of the claim. A self-reported $600 million, inflated 140-fold, flowed into an industry study and out into years of headlines.

Why 2025 broke the charts

CoinGecko counted 11.6 million failures in 2025 alone — 86.3% of all token deaths since 2021, with 7.7 million in the fourth quarter, about 35% of the five-year total in three months.

The mechanism is issuance cost. Launchpads, pump.fun above all, made minting a token approximately free, and free supply produces disposable output. Most of those tokens were never projects. Counting them as "failures" is technically accurate and analytically close to meaningless — it's like counting abandoned tweets.

The quarter also followed the largest liquidation event crypto has recorded. On 10–11 October 2025, per CoinGecko's own account, over $19 billion in leveraged positions were forcibly closed across more than 1.6 million accounts — "nine times the size of any previous single-day liquidation event" — after a 100% tariff announcement on Chinese imports, with Bitcoin falling from $122,574 to $104,782. Open interest had hit a record $217B.

These two facts sit next to each other in the same quarter. Neither CoinGecko nor CoinDesk claims one caused the other, and neither will we — plenty of secondary write-ups do, which is precisely why it's worth saying out loud that the sources don't.

How to use any of this

If you're citing a dead-coin number, cite the definition with it. "53.2% of GeckoTerminal-listed tokens that traded at least once, data to December 2025" is a sentence you can defend. "Half of all crypto is dead" is not — and given that CoinGecko's own two figures imply 66.4%, the safest thing you can do with that headline is not repeat it.

Cite this article