A wallet's profit does not tell you whether a human made it. Its timing does. Bots, snipers and humans leave different marks on the same numbers, and once you know the marks you can read them off a scan card in a few seconds.
On 16 September 2026 we scanned a wallet with 758 trades in 30 days and an average hold of 2h 13m: a human rhythm, no sniper or bundler tag, followed by 13 traders on the data provider's watchlists. Another wallet had 586 trades on 211 tokens in the same window with an average hold of 2h 24m and no tag either, and still scored 25 because 72 percent of its tokens closed at a loss [1]. Hold time tells you what the wallet is. It does not tell you whether it wins.
Why it matters for copy trading
You can only copy a wallet whose edge survives a delay. Your bot sees the target's transaction, builds its own, and lands it a few hundred milliseconds to a few seconds later. For a human trader holding twenty minutes, that delay costs a little slippage. For a sniper holding twenty seconds, the delay is the whole trade: you buy the top it just sold into.
Copying a bot does not give you the bot's results. It gives you the other side of them.
The four marks
Hold time
The clearest one. Average hold is the mean time between a wallet's buy and its sell on the same token.
- Under 10 seconds: a sniper or an arbitrage bot. Nothing here is copyable.
- 10 seconds to 2 minutes: a fast scalper, almost always automated. Copyable only by another bot with the same infrastructure, and even then you lose the spread.
- 2 to 20 minutes: a fast human with a terminal, or a slower bot. Borderline. Look at the other marks.
- 20 minutes and up: a human decision, or a bot set to follow a strategy a human could follow. This is the copyable range.
Block position
A sniper buys in the first block a token exists, often in the same transaction bundle as the launch itself. On a scan card this shows as the sniper tag, and on a token's trader list it shows as a cluster of wallets that all bought in block zero or one.
A human cannot be in block zero. If a wallet's best trades all start in the first two blocks of a launch, its win rate is a measure of its latency, not of its judgement.
Bundle size
A bundler is one operator buying with many wallets in one bundle, so the launch looks like organic demand. The wallets are funded from the same relay minutes before, buy in the same block, and often sell into the buyers they attracted. Each wallet on its own looks like a decent trader. Together they are a single trade.
The funding trail exposes this in one hop: twenty wallets, one funder.
Cadence
Humans have a day. They trade in bursts, sleep, come back. Bots trade every hour of every day at a steady rate. Look at trades over 7 days next to trades over 30 days: a wallet with 400 trades a week spread evenly across all hours is a program.
Cadence also shows in the token count. A human trades a handful of tokens a day. A wallet touching 60 different tokens a day is scanning, not choosing.
The tags on a scan
Ocinct marks five behaviours a follower cannot reproduce, and each one costs 20 points on the score, up to 40:
| Tag | What it means |
|---|---|
| Sniper | Buys in the first blocks of launches |
| Bundler | Many wallets buying in one bundle, one operator |
| Rat trader | Trades against the buyers it brings in |
| Dev team | Trades tokens it deployed, or that its funder deployed |
| Sandwich bot | Front runs and back runs other people's swaps |
The tags come from on-chain behaviour and from public labels. They are a flag, not a verdict, which is why the score shows them next to the numbers instead of hiding the wallet.
A human wallet, for contrast
A wallet worth copying tends to look like this: average hold in the tens of minutes to hours, a few to a couple of dozen tokens a week, trades clustered in a daily rhythm, no tags, funded from an exchange, realized profit larger than unrealized. Its win rate is rarely above 70 percent, and that is fine. See why a 90 percent win rate is a warning sign.
None of those marks alone proves anything. Together, they tell you whether the wallet's edge is a decision you could copy or a speed you cannot.
Sources
[1] Ocinct scan cards, 16 September 2026, risk tags and hold time as computed by the bot: https://ocinct.com/wallet-scanner
[2] Ocinct score method, penalties for sniper, bundler, rat trader, dev team and sandwich bot: https://ocinct.com/how-it-works