$91 billion of fake trades on Base. One wallet, six transactions, less than a dollar.

01 · A Saturday on BaseOne transaction, $36 billion of trades

On Saturday, August 15, one transaction on Base recorded $36 billion of trades. That is about seven weeks of real DEX trading on Base, packed into a single block. Yet it moved almost no real money. Less than a hundredth of a cent in USDC changed hands, plus about two cents of ETH that went out and came straight back. The $36 billion exists only in its swap records.

The wallet that sent it did this six times in August. Together those transactions recorded $91 billion of trades on Base, and the wallet paid less than a dollar for all of them. On the same Saturday it did it once more on Arbitrum.

This is wash trading, trading with yourself so a market looks busier than it is. It is as old as markets. What is new is the price. On Uniswap V4, a wallet can trade hundreds of millions of dollars it never holds, as long as the books balance when the transaction ends. Here they balanced at a hundredth of a cent.

Bitquery indexes every swap on Base, Arbitrum and Optimism. So we went through all of them from January 1 to September 20 and looked for trades that moved almost no real money. We found $108 billion on Base and Arbitrum, made by five wallets, and none on Optimism. Two of the best-known DEX trackers treated those trades in opposite ways.

These totals are a floor. We counted one kind of fake trading, the kind that needs no money, on three chains.

02 · Why fake itVolume is the number everyone checks first

A DEX's volume is the dollar value of every swap made on it. Trackers rank chains, exchanges, pools and tokens by it, and traders read a big number as a sign that a token is popular and easy to get in and out of. For a new token, a spot near the top of a volume list is how it gets noticed.

That makes volume worth faking. The trick is to buy, sell straight back and repeat. Nothing changes hands in the end, but every swap is recorded and counted. On older DEXs this costs real money. Each swap pays the pool's fee, and you need the tokens to swap in the first place. Faking $1 million of volume in a pool with a 0.3% fee burns $3,000.

We have tracked this before, from two wallets that made half the DEX volume on Circle's new Arc chain to fake crowds on Solana. Whether it pays off is another matter. On Solana, tokens that faked volume with a handful of wallets did not pull in real buyers, as our study of wash trading and bots found. On Base, the cost of doing it has all but vanished.

03 · The trickHow to trade $500 million you do not have

Uniswap V4 is the newest version of the largest DEX. Older versions gave every pool its own contract and moved tokens on every swap.

V4 keeps all its pools inside one contract, the PoolManager. During a transaction it keeps a running tab of what each trader owes and is owed, and tokens move only once, at the very end, to settle whatever difference is left. Uniswap calls this flash accounting. It saves real traders gas when a trade hops through several pools.

It also means the tab can be enormous for a moment, as long as it comes back to zero. Only the final balance is checked. On Tuesday, August 18, one transaction from the same wallet did five things in a row.

Every one of those 510 swaps really happened, and the chain recorded each at full size. The last step cancels the debt. Liquidity is the pile of tokens a pool trades against, and adding $500 million of it leaves the wallet owing the pool $500 million. Selling its own new token into that pool pays the $500 million back out. The debt and the payout cancel, and the tab closes at 0.000101 USDC.

We rebuilt the whole transaction from its events with Uniswap's own math. What the events say the wallet owed matches what it paid, to the last unit. The transaction recorded $9.6 billion of trades.

None of this needed a hook, the add-on code a V4 pool can run. Four of the eight pools named the contract the wallet called as their hook, but V4's code skips a hook when the hook is the one calling. Hooks can hide things in other ways, as we found in a pool that shows a 0% fee and takes 1%.

04 · How big it gotMost of Base's August volume was fake

On Base the fake trades came in two bursts. In January, one wallet wash-traded a new token called ROLL for five weeks, about a quarter of the DEX volume we recorded that month. In August, the wallet from the opening made $91 billion in six transactions. Then it went quiet.

That month the fakes were 80% of all DEX volume we recorded on Base, almost four times the real trading. Arbitrum had one such transaction, from the same wallet on August 15. It came to $5.5 billion, more than half of what we recorded on Arbitrum that month. We found none on Optimism, even though bots make most of its trades, as our look at who still trades on Optimism showed. Almost all of it traces back to two wallets.

05 · Who did itTwo wallets, almost all of it

Five wallets made every fake trade we found. Two of them made 99.99% of the dollars, and they could hardly be more different.

The August wallet is hard to explain as marketing. All 14 tokens it traded were created inside its own transactions. Nobody else ever traded them. Each pool lived for one transaction. Its first swaps use round sizes, such as $31.25 million, a sixteenth of the $500 million. That reads like a script. It could be testing, or padding tracker data. The chain does not say which.

The January wallet looks like the classic reason for wash trading. ROLL was deployed on January 14. Its first recorded trades, the wash wallet's first ETH and the wash trading itself all began on January 16. Over five weeks the wallet sent 103,392 transactions, nearly all of them running 50 swaps back and forth in the ROLL/USDC pool through Uniswap's own router, and not one of them moved any USDC.

Other wallets traded $193 million of ROLL by September 20, from more than 24,000 addresses. No money moved directly between ROLL's deployer and the wallets behind the wash trading, so who paid for it is unknown. The funding trail can be followed with the Bitquery MCP.

Three smaller wallets did the same with CGN, LODE and a memecoin called Basecat, from about $45,000 to $9 million each. None of the five paid much for the privilege.

06 · The billLess than a dollar for $91 billion

Faking volume the old way means paying a pool fee on every swap. These wallets paid almost nothing.

The August wallet spent $0.43 on Base for $91 billion of trades, counting gas and every real token that left the wallet for good. The ROLL wallet sent far more transactions and paid about $2,400. Run through a pool that charges a 0.3% fee, the August wallet's volume would have cost about $274 million in fees alone. The trackers then had to decide whether to count it.

07 · Who counted itGeckoTerminal counted it. DefiLlama did not.

GeckoTerminal, CoinGecko's DEX tracker, counted the fake trades in its public data. For one of the August 18 pools it reports $1.98 billion of volume that day. Its API still returns a pool from August 15, RS/USDC, with $9.1 billion in reserves, although its website has no page for it. That pool holds almost no USDC. The ROLL/USDC pool does have a page, and GeckoTerminal's figures for it match the January volume we counted.

DefiLlama did not count them. On August 15 it showed less Uniswap V4 trading on Base than on a normal day. DEX Screener never listed any of the pools we checked.

Take the fakes out and our totals for Base and Arbitrum land within 3% of DefiLlama's for the year, which fits a tracker that never counted them. Bitquery's own DEX data records these swaps like any other trade, and so does any feed that reads swap events straight from the chain, including our Uniswap API. The figures here come from pulling them apart.

08 · How to spot itFour signs a volume number is fake

Every case we found shared a few tells, and each one can be checked from public data.

• One wallet makes almost all the volume. Real pools have many traders. These had one, and a pool's top traders show it at once.
• The pool is brand new. The August pools were created and used up in the same transaction.
• Billions trade and the price does not move. A round trip puts the price back where it started.
• Trackers disagree wildly. When one shows billions and another shows almost nothing, look at who traded.

09 · Does it matter?It depends on who reads the number

Chain rankings built on DefiLlama's figures left these trades out. Anyone reading volume straight from swap events took them in, and across August, Base looked about five times busier than it was. A new token can look like a top pool for the price of gas.

So the fix sits with the reader. Before trusting a big number, check who made it, especially on a pool that did not exist yesterday. Bitquery's DEX data shows both, trade by trade.

A volume figure says how much was traded. On Base on August 15, the answer to who traded it was one wallet.

10 · How we measuredWhat counted, what did not, and why

We counted a transaction as wash trading that needed no money when two things held. First, its swaps went both ways through one pool, or they traded only against liquidity the same transaction had just added. Second, the real tokens that moved, meaning stablecoins, ETH and bitcoin, came to 1% or less of the trades' dollar value, with at least $10,000 traded. Only swaps in pools that passed the test count toward the totals.

This is a floor. It misses wash trading spread over several transactions, round trips on older Uniswap versions, which move real tokens, and pools where neither token has a dollar price. On Base before May 31 we could only check the first pattern. We checked 150 flagged transactions by hand, and all were wash trading. We also checked the 11 near misses, trades that almost qualified, and none was. Two separate readings of the chain, one from decoded trades and one from raw pool events, agree on the August wallet to the cent.

11 · The recordThe addresses behind the story