Ten wallets that buy the same token in the same second are not ten traders who agree. They are one keyboard. On a leaderboard they look like ten independent winners, and that is exactly the point of running them. This guide is about spotting the set before you copy any of it.
The thresholds in this guide are the ones Ocinct runs in production since September 2026: buys in the same slot, about 316 milliseconds as measured on the chain, are the same hand, a leader has to be first on half the shared tokens and clearly ahead of the second, and 3 shared tokens is the minimum before the tool says anything at all [1].
What a crew is
A crew is a group of wallets operated together. The reasons vary: to make a launch look like demand, to hold more of a supply than one wallet could without drawing attention, to place several wallets on a leaderboard so that one of them gets copied, or simply because the operator rotates wallets and keeps a few alive at once.
Whatever the reason, a crew has three properties you can measure.
The three signals
A shared funder
Crews are set up in a batch. One relay wallet sends 1 to 3 SOL to each member within minutes. Open the Funded by line on each wallet and the same address comes back. This is the cheapest check and it catches most of the naive crews. The funding trail shows it in one hop; how to read the label covers the details.
Sophisticated operators fund each wallet from a different exchange withdrawal. That defeats the funder check and leaves the next two.
Same second buys
Two humans do not buy the same token within one second of each other more than once by accident. One script does it every time. Pull the recent activity of the suspect wallets, line up the timestamps on shared tokens, and count how often the gap is under a second.
Ocinct orders the buys by slot rather than by timestamp. A provider rounds to the second, a slot lasts about 316 milliseconds when measured in September 2026, so two buys in the same second can still be ordered. Two buys in the same slot cannot, and the tool says so instead of inventing a leader: repeated across at least three tokens, that is one operator.
A stable leader
The softer version is a follower, not a twin. Wallet B buys 10 to 90 seconds after wallet A, on token after token, and never before. That is a copy bot, or the operator's own second wallet firing after the first. Either way, copying B means paying for A's signal twice.
Ocinct measures the share of first places across shared tokens, not the reaction time. A crew enters minutes apart, so the leader is the wallet that arrives first again and again: first on half the shared tokens, and at least one and a half times more often than the runner up. Measured on a real group of three in September 2026, the leader was first on 61 percent of the tokens all three bought and the next on 21 percent. Anything closer than that is called no leader.
How crews inflate leaderboards
Run twenty wallets on the same launches. Sell most of them into the buyers you attracted. Some wallets will end up with bad records and some with great ones; that is variance, not skill. Now the top three are on a leaderboard sorted by PnL, and nobody sorts by funder.
Copy one of the three and you copy a coin flip with the losing side hidden. Worse, on the next launch the crew's sells are the ones you buy from.
Keeping crews off your list
- Scan the list, not the wallet. When you pull the top traders of a token, look at the funder column before the profit column. Group by funder. Anything with three or more members is one entry, and usually not a good one.
- Check timing on the survivors. For the wallets that pass your filters, run who copies who or compare their recent buys by hand. Same slot buys or a fixed leader means drop the followers.
- Distrust perfect sets. Five wallets with 80 percent win rates that all trade the same twelve tokens are a set. Real independent traders overlap on the big tokens and diverge on the rest.
What the tools do
On Pro and Elite, the token traders tool returns the traders of any token with their profit, tags and funder, and groups the ones that share a funder. Who copies who takes one wallet or a group of up to ten and reads the chain: same hand pairs, who buys first, and the wallet outside your list that all of them follow. A name is only printed when it beats a control window placed five minutes earlier, so a crowd of snipers is never mistaken for a source. Both run on tool credits and refund a failed run.
By hand it is slower but the same work: funder first, timestamps second, leader third.
Sources
[1] Ocinct, who copies who, thresholds as coded in the bot: https://ocinct.com/who-copies-who
[2] Helius, Wallet API, Funded by, wallet clusters and Sybil detection: https://www.helius.dev/docs/wallet-api/funded-by