How we know a buy is a buy

The whole copy book rests on one question: did this trader actually buy this token? On Robinhood Chain that question is harder than it looks, and a lot of people are working to make it look wrong. This chapter is the concrete version: what a real buy looks like on the chain, what the fakes look like, and how each one is filed.

Everything below was measured on the tracked wallets in the first week of September 2026. The numbers are ours.

The accounts are not ordinary wallets

The traders fomo ranks do not use plain wallets. Each account is an ERC-4337 smart account: a contract wallet whose owner signs UserOperations that a third-party bundler packs into a transaction and submits to an EntryPoint contract. Robinhood Chain runs three EntryPoints (v0.6, v0.7 and v0.8); we read the event, not the address. The transaction's sender is the bundler, never the trader — so the naive rule, "the trader signed it", finds nothing. Over 30 days the 69 main accounts produced 12,475 UserOperations, and every one of them was a sell, an approval or a transfer. They never buy through their own account.

How a real buy arrives

fomo executes a buy through Relay (relay.link), a cross-chain execution network. The trader pays from their fomo balance — on Solana, Base, BSC or Robinhood Chain itself — and a Relay solver submits the transaction here: solver → Relay's approval proxy (0xccc88a9d…) or router (0xb92fe925…) → the pool → the trader's account. The transaction is signed by the solver, the pool is a normal Uniswap or launchpad pool, and the token hops through the router and proxy before it lands.

Relay publishes a record of every request it filled (GET /requests), and that record is where the truth lives. A request created by fomo's app is tagged referrer: "fomo" and carries fomo's app fee to fomo's fee wallets. In the week we measured, 1,536 of 1,622 Relay fills landing on tracked accounts carried that tag.

So a buy is three facts, on any venue, and all three must hold:

1. The wallet authorized it. Any one of: it signed the transaction; its UserOperation is in the receipt; its account executed a call inside the transaction (read from the call trace, which covers session keys, delegated executors and whatever account flow comes next); its own WETH, USDG or the pair's quote token left it in that transaction, which only its approval allows; fomo created the Relay request; or the wallet verifiably funded a Relay request on the origin chain — Base, BNB or Solana, read there directly.
2. Its money moved. Native ETH out of the wallet, or one of those paying assets, or Relay's recorded input. An authorized action that paid nothing is a claim — a mint, an airdrop the account pulled itself — and is filed as such, never copied.
3. The receipt is honest. The token really arrived, it reached a plausible number of final holders rather than a crowd, and a swap event counts as evidence only when a contract other than the token emitted it — a token can emit any log it likes about itself.

Nothing in that list names a DEX. Uniswap v2, v3 and v4, Pancake, Ramses, Giga, Ekubo, Orvex, Bankr, o1, the Pons and Flap curves — the rule reads the same facts on every one of the forty venues on the chain, and the swap-event list only sharpens the cost estimate.

The Relay record also tells us what the trader paid, which is the copy's size reference, and the buy's age is exact: a buy first seen long after the fact is recorded but never copied. Sells obey the same authorization test: an outbound transfer the trader did not authorize — a zero-value poisoning transfer, a fake token's log — is not a sell.

What the fakes look like

Everything else that lands tokens in a famous account is a delivery, filed as a deposit with its reason, never a buy. Five shapes, all seen this week:

The spray. An airdrop contract or Multicall3 pushes a token to 30 tracked traders in one transaction. Cheap, obvious, and until we counted final holders it rode into the book on the back of any swap in the same receipt. One transaction, dozens of holders: not a buy.

The buy-for-you. A stranger's wallet calls a Uniswap router's swapExactETHForTokens with to = the trader. Real money, one recipient, a real swap — and fomo lists it as a deposit, because the trader did nothing. The most expensive one we saw put 0.47 ETH of a token into unipcs' account. Filed by rule 1: the trader neither signed nor carried a UserOperation.

The Relay spray. The same idea through Relay's own website: one payer, referrer: "relay.link", the same token for six traders in twelve minutes at $200 each, or for thirteen traders at fifty cents. Filed by the payer check — one payer, many tracked recipients — and by the missing fomo tag.

The written-in requester. The subtle one. Relay's request record has a user field, and it is whatever the requester typed in. A pool wallet on Base funded buys for top traders for days with each trader's own address written into user, so the record read "unipcs bought this, paid by unipcs". On the chain the Base deposit came from the pool, submitted by a rotating operator wallet, and the request carried no referrer — fomo's app had nothing to do with it. This one passed every check that looked at the payer alone; the fomo tag is what stops it.

The stale buy. Not a fake, but the same damage: a genuine buy surfaced hours late by a lagging poll, copied at whatever the price had become. The freshness rule filters it.

The other guards

• Final holders, not recipients. A Relay fill passes the token through two contracts before the trader; a distribution passes it to a crowd. We count the addresses that keep it.
• Payer exclusivity. A payer funding fills for more than one tracked account is not a trader, whatever the tag says.
• Quote sanity. A chain quote far away from the market price is the wrong pool or the wrong direction, never a price; a paper entry takes the market price instead.
• Fresh launches only for side wallets. Who bought seconds before a trader means something at a launch and nothing on a month-old pair.

The thresholds themselves

Each guard above is a number — how fresh, how far, how many. This chapter names the mechanisms, because they are what makes the book worth trusting and every one of them is checkable against the chain yourself. The thresholds stay ours: the adversary in the next section reads the docs too, and a published cut-off is a cut-off you size a spray to sit under.

Why it matters beyond copy trading

Every copy bot that follows fomo's leaderboard is paying for the same fakes. A spray costs the attacker a few dollars and produces, across every follower, real buys at the top of a token they are about to dump. The published research on meme-coin copy trading describes exactly this adversary — coordinated wallets timed against copiers — and reaches the same conclusion we did: trust only the actions the platform itself executed for the user.

Further reading

• Relay: Get Requests — the request record, including user, recipient, referrer and the fee fields.
• ERC-4337: Account Abstraction — why the trader's signature is not on the transaction.
• Resisting Manipulative Bots in Meme Coin Copy Trading — sniper, bump and spray adversaries against copy traders, measured.
• fomo — the app: one USD balance across Solana, Robinhood Chain, Base, BSC, Ethereum and Monad, a 0.5% fee.
• Robinhood Chain explorer — read any transaction mentioned here yourself.