Most copy traders pick a wallet the way they pick a token: someone posted it, it looked good, they pasted it into a bot. Then they lose money on a wallet that was never copyable in the first place. This guide is the other way round. Start from a pile of candidates, remove the ones that cannot work, and keep the few that can.
We ran this method on 16 September 2026 with the Ocinct bot on two wallets pulled from a token's top traders. One scored 82 of 100: $15k realized in 30 days on 181 tokens, 70.8 percent sold in profit, average hold 2h 13m, funded 34 days earlier with 0.041 SOL. The other scored 25: $32 realized on 211 tokens, 22 percent sold in profit. The Steady winners preset kept the first and dropped the second [1].
Where candidate wallets come from
There is no shortage of addresses. There is a shortage of vetted ones. The sources below are ranked by how much noise they carry.
- The top traders of a token you watched pump. Open the token on an explorer or a terminal, sort holders and traders by realized profit, and copy the top 30 to 50 addresses. This is the best source because it is anchored to a real event and not to someone's opinion.
- Public leaderboards. Ranked by PnL, refreshed daily, easy to scrape. Also the noisiest: leaderboards reward snipers, bundlers and insiders, because those make the most money on paper.
- Telegram channels and X accounts that post wallets. Useful for volume, useless as a filter. Treat every posted wallet as a candidate, never as a pick.
- Wallets you copied before. Old picks that went quiet often come back. Rescan them monthly.
Aim for a list of 50 to 100 addresses. Fewer, and the filters leave you with nothing. More, and you will not read the survivors.
The five checks, in order
Run them in this order because each one is cheaper than the next, and each one removes a different kind of wallet.
1. Is the profit realized?
A wallet shows two profits: what it took out (realized) and what its open positions are worth (unrealized). Leaderboards often quote the sum. Only the realized part is money.
A wallet with $40,000 realized and $90,000 unrealized is holding bags it cannot sell without moving the price. Copy it and you hold the same bags, except you bought later. Keep wallets whose realized profit is most of the total.
2. Is the win rate earned?
Win rate is the share of tokens closed in profit. Two things fake it.
The first is selling early and never closing losers: a wallet takes 5 percent on twenty tokens and keeps the five that went to zero open. That is a 100 percent win rate on paper and a loss in reality. The outcome buckets show it: many small wins, nothing above 2x, and a stack of unrealized losses.
The second is a structural edge you cannot copy: sniping the first block of a launch, bundling, trading its own launches, sandwiching. Those wallets carry risk tags on a scan, and their win rate comes with them. See how to tell if a wallet is a bot.
3. Is it still trading?
A wallet that made $30,000 three weeks ago and nothing since is not a strategy, it is a story. Check trades in the last 7 days and the time since the last swap. A good month with a dead week usually means the operator moved to a new wallet.
4. Can you copy the hold time?
Average hold under a minute means the wallet's edge is speed. It buys in the same block as the news and sells before you have seen the token. You will buy after it and sell after it, on both sides of the spread. Holds of 20 minutes to a few hours are the range where a human with a bot can keep up.
5. Who funded it?
The first SOL a wallet received came from somewhere, and explorers show it as Funded by [2]. If it came from an exchange, the wallet is probably a person. If it came from a wallet that also funded twenty other top traders, you are looking at one operator with twenty wallets, and the top three of them are on the leaderboard by selection, not by skill. See what Funded by means.
Doing it in one pass
By hand, the five checks take five to ten minutes per wallet. On a list of 100 that is a weekend.
Ocinct does the same five checks in one list scan. Paste the addresses, apply a preset or your own rules, and get back the wallets that pass with a CSV of every result. The score sorts the list; the fields on each card show why a wallet passed. The Free plan scans three wallets a day, which is enough to check the ones a channel posted. The list scan is on Pro and Elite.
A reasonable first filter for copy trading by hand:
| Rule | Value | Why |
|---|---|---|
| Tokens sold in profit, 30d | 50 percent or more | Below that the wallet needs big wins you cannot time |
| Profit taken, 30d | $2,000 or more | Enough to be a strategy, not luck |
| Profit taken, 7d | 0 or more | Still working this week |
| Average hold | 20 minutes or more | Copyable without being first |
| Last trade | 48 hours or less | Still alive |
| Tags | not sniper, not bundler | Edge you cannot copy |
Expect 100 wallets to become 5 to 15. Then read those cards.
What to do with the survivors
Copy small. Copy bots let you set a fixed buy size per wallet you follow [3]. A wallet's average buy size is on its card; start with a fraction of it and raise it after two weeks of watching your fills next to its fills. Cap the number of wallets you copy at once: three good ones beat ten average ones because each copy adds slippage and your bot cannot fire ten at once.
And rescan. A wallet that passed in September can be flagged in October. Put the survivors back through the list scan every couple of weeks and drop the ones that stop passing.
Sources
[1] Ocinct scan cards, 16 September 2026, computed with the bot's published score method: https://ocinct.com/how-it-works
[2] Helius, Wallet API, Funded by: https://www.helius.dev/docs/wallet-api/funded-by
[3] Bloom docs, copy trading presets: https://docs.bloombot.app/solana/solana-bot/settings/presets/copy-presets