Beyond "Crime Tokens"

I started this research because “crime token” had become one of those phrases that explains everything and therefore explains nothing. A token launches, collapses, and the label gets attached afterwards. Sometimes people mean bad tokenomics. Sometimes they mean a coordinated pump. Sometimes they just mean they lost money.

So I pulled apart 22 tokens that had already suffered extreme drawdowns and sorted them by a simple question: when did the highest price happen?

That one column changed the whole picture.

The tokens did not form one family. They split into two. One group peaked at launch and bled for months or years. The other sat almost untouched for months, went vertical in days, and then gave the entire move back just as quickly.

Two different machines can produce the same wreckage.

The short answer: these 22 failures split into two mechanisms. Low-float launches often peaked while public demand was highest and tradeable supply was lowest. Quiet tokens failed differently: leverage accumulated first, then a short squeeze created a violent round trip. The same collapsed chart can hide very different market structures.

The slow unwind starts on day one

The first archetype is the low-float unwind. A small fraction of supply becomes tradeable at launch, while the team and early investors keep large locked allocations that vest over several years.

On listing day, attention is at its maximum and available supply is at its minimum. Exchange promotions, airdrop claims and launchpool rewards all arrive together. The opening price looks like a market price, but it is being formed in the least representative market the token will ever have.

After that, both sides move the wrong way. Attention fades. More supply unlocks. Early recipients with a zero or near-zero cost basis sell into a thinner bid. Nobody has to break a rule for the chart to grind lower. The design can do the damage by itself.

The numbers behind that design are enormous. Binance Research estimated that roughly $155 billion of tokens were scheduled to unlock between 2024 and 2030. Holding prices flat would require about $80 billion of new demand. At the same time, launchpool and airdrop tranches put zero-cost tokens directly into exchange accounts on the day when promotional demand was hottest.

0 days: 6 tokens. 1–2 days: 4 tokens. 8–21 days: 4 tokens. 77+ days: 4 tokens.

In six of the 18 tokens in this group, the all-time high arrived on listing day. Ten peaked within 48 hours. Fourteen peaked within three weeks.

That is the part I keep coming back to: for most of these assets, the highest price ever available to the public was also the first one. The chart was not discovering value. It was clearing a temporary imbalance between maximum demand and minimum supply.

The final drawdowns are brutal, but the path is usually slow. The buyer harmed most is the person who mistakes a launch valuation for a durable one and keeps averaging into a two-year vesting schedule.

The schedule was part of the price

Five of the 22 projects changed material token parameters after launch. That deserves more attention than it gets, because it means buyers were not only pricing a difficult deal. In some cases they were pricing a deal that could be rewritten after they entered it.

STRK supplied the cleanest experiment. Its original April 2024 cliff would have released 1.34 billion tokens, roughly 13% of the total supply, in one day. After backlash, StarkWare cut the release to 0.64%. The price rose more than 20% on the announcement.

That move answered the argument by itself: the market was pricing the schedule before the new supply existed. The team made the most holder-friendly intervention in the cohort, and STRK still ended up 99.3% below its high.

The vesting shape mattered too. Cliff unlocks created dated shocks that traders could front-run. Linear vesting replaced the event with permanent pressure. In Keyrock's aggregate study of more than 16,000 unlocks, roughly 90% were net negative for price, with weakness often beginning before the unlock date and taking as long as 30 days to play out.

The fast collapse needs a different machine

The second archetype looks nothing like that.

These tokens can trade flat and thin for months. Then the spot price begins to rise, perpetual volume expands, shorts pile in, and liquidations start funding the next leg upward. If a small group effectively controls most of the tradeable supply, it does not take much spot buying to move the mark price. The derivatives market does the rest.

RAVE is the cleanest visual example in the research.

Start

$0.239

All-time high

$28.10

Latest

$0.292

From high

-99%

The series starts at $0.239, reaches a high of $28.10, and ends at $0.292.

The move from roughly $0.25 to around $28 took only days. Most of it disappeared over the next 24 hours. Reporting attributed roughly $44 million of liquidations during the rally, mostly to short positions.

MYX showed the same distortion in a different form: $6–9 billion of daily perpetual volume around a token whose circulating value in the research snapshot was only $13.3 million. More than $10 million of shorts were liquidated in a single day. Calling that “liquidity” misses what the number means. The derivative had become larger than the market it was supposed to track.

This is not the same risk as a launch-day peak. The slow unwind hurts longs through dilution and fading demand. The squeeze can hurt shorts on the way up and late longs on the way down. One is mostly a cap-table problem. The other requires a particular market structure: concentrated effective supply, a small spot float, and a perpetual market large enough to dominate price discovery.

There are serious manipulation allegations around several tokens in this second group. They should remain allegations unless wallet ownership and intent can be proved. Exchange custody is especially easy to misread: a large omnibus wallet shows where customer tokens sit, not who owns them.

The distinction also appeared in my risk framework. LAB and SIREN scored 9.4 out of 10, RAVE 9.3, RIVER 9.0 and MYX 8.6. All five post-listing squeeze tokens occupied the top of the table because of concentration and transparency, not because a score predicted their returns. EIGEN scored only 6.1 and still lost 96.4%. The score was useful for identifying the machine, not forecasting the wreck.

The control case broke my first thesis

At first, the obvious explanation seemed to be low float and high fully diluted valuation. Nineteen of the 22 collapsed tokens had less than 20% of supply circulating at launch. The 2024 launch cohort averaged roughly 12.3% market cap to FDV in Binance Research’s study.

Then HYPE broke the story.

Hyperliquid launched with a float that would trigger the usual warning screen, yet its outcome was nowhere near the cohort’s median drawdown. The difference was not the percentage locked. It was the identity and incentives of the holders behind the lock. HYPE had no VC allocation and no private sale. There was no large investor tranche with a near-zero cost basis and a fund lifecycle eventually forcing distribution.

It also had something most of the collapsed cohort did not: meaningful product revenue connected to the token’s ecosystem.

That changed the question for me. “How much is locked?” is too crude. Who owns the locked supply, what did they pay, and why will they eventually sell? is much closer to the mechanism.

The control is even sharper beside SAGA. HYPE had about 23% circulating and was down roughly 26%. SAGA launched with about 9% circulating and was down 99.8%. Low float described both. It explained neither. What separated them was the cap table and whether the underlying product generated real, token-connected revenue.

TVL was another false comfort

The same problem appears in valuation ratios. EIGEN looked extraordinarily cheap on FDV divided by TVL: about $376 million of FDV against $5.2 billion of TVL in the research snapshot. It was still down more than 96% from its high.

The ratio failed because restaked ETH is not revenue accruing to token holders. A large denominator can make a valuation multiple look sensible without creating a single dollar of recurring demand for the token.

BLAST made the same point from below one times TVL. Its FDV was only 0.88 times TVL, it allocated 50% of supply to the community, and it still fell 99.1%. “Community allocation” did not make mercenary recipients patient, and a cheap-looking ratio did not turn activity into cash flow.

ZORA showed what happens when the product story disappears. Its creator-coin thesis was retired by Base within ten months of the token launch. By the research snapshot, Zora chain TVL was $36,542 while the token still carried a $63 million FDV. The token had outlived the reason people were told to value it.

The more useful denominator is revenue that actually accrues to the asset. If that number is approximately zero, a low FDV/TVL ratio is decoration, not support.

A chart cannot prove intent

MOVE is the case that made me most cautious about storytelling from price alone.

Its chart resembles the ordinary low-float unwind: the token peaked a day after listing and then fell. But public reporting later documented a market-making arrangement that allowed 66 million MOVE to be sold for roughly $38 million. Binance offboarded the market maker, Coinbase delisted the token, and Movement Labs eventually filed for Chapter 11.

The important lesson is uncomfortable. A chart with a documented market-maker dump can look identical to one produced by many independent holders responding to the same incentives. Price shape can tell us which mechanism is possible. It cannot tell us who intended what.

That cuts both ways. The shape is not proof of misconduct, and a smooth-looking decline is not proof that nothing improper happened.

The screen I use now

I no longer start with float. I start with the cap table: who owns the locked supply, their cost basis, and the exact path by which it becomes liquid. Then I look at whether the product produces revenue that reaches the token, rather than TVL or activity that merely sits nearby.

For a new launch, I treat the opening price as an ask, not as established value. If the perpetual market is already large while spot depth is tiny, I treat that as leverage risk rather than healthy liquidity. If tokenomics change after launch, I assume the original deal was mispriced until proven otherwise.

And when a quiet token suddenly rises ten or one hundred times months after listing, I do not confuse that with the slow unwind. It is a different machine, with a different victim and a much shorter clock.

The phrase “crime token” may survive because it is memorable. As analysis, it is too blunt. The useful distinction is simpler: was this asset designed to leak value slowly, or was its market structured so value could move violently?

That question does not produce a verdict. It does produce a better investigation.