Base wash trading: two wallets bought $2.2 billion of tokens from their own pools.

01 · A Tuesday on Base$199,000 in, $199,000 back

On the morning of Tuesday, September 15, a wallet on Base spent $199,000 of USDC on about two billion units of a token called POOLS, in an Aerodrome pool. Six seconds later the pool sent the same $199,000 back to the same wallet.

Then it did it again, and again, a buy about every ten seconds. Eight minutes later it had bought 46 times. The pool had sent back 46 payments that came to five dollars more in USDC than the wallet had put in.

POOLS showed a market cap of about $10 million at the time. Almost all the trading in it that month, 99.99% by dollars, was this one wallet.

02 · The trickBuying from a pool you filled yourself

A DEX pool holds two assets, here USDC and a token. The money in it comes from liquidity providers, who earn the fees traders pay. When someone buys the token, their USDC goes into the pool.

If the buyer is also the pool's only liquidity provider, that USDC is already theirs. They can collect it straight back, fee included. The trade still counts as volume, the dollar figure trackers rank tokens by, and it costs almost nothing.

Two Base wallets did this for the whole month to September 22. They bought $2.18 billion of tokens, and across the month each got back more than 99.9% of the USDC it sent to those pools. With three bots and two helper wallets tied to them by transfers, the seven made seven in ten dollars of all the trading our tests flag as not real on Base.

03 · The pairTwo wallets, one after the other

The two buying wallets worked in turn. One did its buying until September 3, the other from that day to the end of our window. Across the month they bought far more than they sold, because the money they got back came mostly from their own pool positions rather than from sales. In the pools we opened, the wallet had put the tokens in itself before its first buy. For the rest, the USDC coming back is what shows it.

Each buying wallet had a helper that passed USDC back and forth with it. On September 3, as one buying wallet stopped and the other started, one helper sent $212,496 to the other. Three bots traded the same tokens in the same pools. Every one of the seven sent USDC to at least one of the others or got it from one.

One buying wallet handed its helper tokens by the tens of billions of units. The helpers did almost no buying of their own, and their sales brought in about $104,000 more than their purchases, between them.

04 · The tokens111 tokens, one address ending

The 111 tokens behind the pair's own-pool buying all have contract addresses ending in b07, a suffix thousands of Base tokens share, so it tells us nothing about who made them. They carry about fifty different names. Sixteen are named OpenAI, twelve LAPTOP, two Anthropic and two PEPE. Anyone can give a token any name. None of these is issued by OpenAI or Anthropic, and none is the well-known token of the same name. Our LAPTOP story, corrected on September 24, shows one of these wallets doing the same on a LAPTOP copy.

Almost no outside money touched them. The seven wallets made all but, at most, about $500,000 of the trading in these tokens, and by our buyer test a few hundred other wallets bought about $179,000 of them. A reader who saw the volume on a tracker saw the pair trading with itself.

05 · The serviceA second machine: 23,641 wallets, one contract

A separate operation ran on Base in the same weeks. For two weeks in September, 23,641 wallets sent transactions to one contract, across 3,120 new tokens, each with its own Uniswap pool. The transactions we opened bought and sold one token five times each way. It recorded $785.5 million of trading, about a quarter of Base's not-real total.

In the three transactions we opened, the pools kept almost nothing, at most 0.001% of the ETH, and the ETH went back out to a different wallet from the one that paid it in. All seven of the service's wallets that we checked have pointed their accounts at the contract, through a feature called EIP-7702 that lets an ordinary wallet borrow a contract's code.

The service's tokens also showed crowds of early buyers. In the 3,004 Base tokens pushed by one volume service, about 21 wallets bought each token in its first four minutes; in the two launches we traced, all 40 had been sent ETH seconds before their first buy, in transactions signed by one address per launch. Those are the tokens where we could see who bought first.

06 · What we cannot sayThe limits

• Who runs the wallets or the service. We name addresses only. Nothing we found ties the service to the seven wallets.
• Where the tokens came from. We could not find a creation record for the tokens we checked, so we name no launch tool.
• Whether outsiders lost money. Almost no one else traded the pair's tokens, and we have no fair comparison for the service's tokens.
• Why. A volume ranking would fit, and nothing on the chain proves it.

07 · QuestionsWhat readers ask

Why would anyone buy from their own pool? To make a token's volume look big. Trackers and screeners rank tokens by volume, and this kind costs almost nothing.

Does it cost anything? Very little. The fee goes back to the only liquidity provider, which is the buyer. In our scene the pool paid the wallet $5.42 more USDC than it sent, before gas.

Is this the same as the trick in your earlier Base story? No. That investigation covered trades that needed almost no money, almost all of them before this window. Here the money is real, and it goes round in a circle.

What is a volume service? A contract that makes round trips for many tokens at once, for whoever sets it up, so each new token shows trading from its first minutes.

08 · How to spot itThree signs of own-pool trading

Each of these can be checked from public data.

• One liquidity provider. A pool whose only liquidity comes from one wallet, and that wallet is also the biggest buyer.
• Money that comes straight back. Each buy followed within seconds by the same USDC paid back to the buyer from the pool.
• Round, repeated clips. The same amount, $199,000 here, every few seconds. A token's trade history shows all three.

09 · How we measuredWhere the numbers come from

We read every DEX trade on Base from August 24 to September 22 that we could price. By our tests, 13.1% of it was not real. The pair's figures count their buys in that window. To check the money coming back, we compared the USDC each wallet sent to its pools with the USDC those pools sent back, from August 20 to September 23. The links between the seven come from USDC they sent each other.

The service's figures count every trade sent to its contract. On September 24 we checked seven wallets around it, the ones that sent its transactions, funded them and received its ETH, and all seven pointed at the contract. We opened a handful of the pair's pools, including the POOLS pool and a LAPTOP copy; in each, the buyer had supplied the pool itself. Our earlier Base investigation counted $46,188 of its kind of trading on Base from September 1 to 20. Its test looked for trades that moved almost no money, and each of the service's transactions moves real ETH, in from one wallet and out to another. Our six-chain overview puts Base beside five other chains. Bitquery's own DEX data records these swaps like any other trade; the figures here come from pulling them apart.

10 · The recordThe addresses behind the story