Iceberg entries, merges, and decoys: how sharp traders evade copy-bots

Copy-bots watch for visible trades — so traders who don't want to be copied have learned to trade invisibly. The three standard evasions are iceberg accumulation (entries sliced too small to trigger alerts), merges (exits that never appear as sales), and wallet rotation (the record you're watching isn't where the action is). Polymarket's own newsletter has described sharp traders doing exactly this, and the techniques matter twice over: they quietly degrade honest copies, and in deliberate hands they're the toolkit that turns copiers into exit liquidity.

Why would a profitable trader hide from copiers?

Because copiers cost them money. A bot mirroring a sharp wallet piles into the same thin order book moments later, pushing the price toward fair value before the sharp can finish building — front-running in reverse, paid for by the leader's remaining edge. A trader whose edge depends on accumulating quietly at good prices has every incentive to make their activity hard to follow. Evasion isn't inherently sinister; it's self-defence. The problem is that the same toolkit, pointed at followers instead of away from them, is a farming operation.

Evasion 1: iceberg entries

Instead of one $20,000 fill that every tracker flags, the position arrives as a hundred small slices — classically under $200 each, sized to duck the copy-bot and whale-alert thresholds — spread across hours or days. By the time the position is visible at full size, the price the leader paid no longer exists; followers chase fills the iceberg already consumed.

For a copier this means the entries your bot can see from such a wallet are systematically the worst ones. When iceberg accumulation precedes an advertised winning streak, it stops being defence and becomes one of the farming signatures we detect: the quiet build is the setup, the publicised record is the bait.

Evasion 2: the merge — an exit with no sale

Polymarket's conditional-token mechanics allow any YES + NO pair of the same market to be merged back into $1 of USDC at any time, with no trade on the order book. A trader holding a large YES who wants out doesn't have to sell — they can buy the matching NO (small, sliced, unremarkable) and merge the pair into cash.

To a bot watching for sells, nothing happened. The leader is flat; the copiers are still long, holding a position whose owner has already left. Used systematically against followers this is the stealth-merge farming pattern — and it's why a bot that only mirrors visible trades, rather than tracking net position deltas, has a structural blind spot you should ask about before choosing one.

Evasion 3: wallet rotation and decoys

The cheapest evasion is to simply not be the wallet everyone watches. Polymarket wallets are free and KYC-less; a trader can run entries from fresh addresses while the famous one idles — and a farmer can go further, parking losses on linked sub-wallets so the flagship's record stays pristine. The wallet you're vetting may be a storefront. This is why our forensics include funding-graph analysis: addresses funded from a common source get judged as a cluster, not as the flattering member alone.

What a copier should take from this

• Assume the visible record is curated. The trades a wallet lets you see are not a random sample, and on this platform a quarter of all volume is estimated to be self-dealing to begin with.
• Prefer wallets whose edge survives being followed. Low-turnover wallets in deep markets lose little to copiers and have little reason to evade; speed-dependent wallets in thin books have every reason — and their edge wasn't copyable anyway.
• Check the forensics, not the PnL. Iceberg builds, merge exits, and decoy structures are invisible in a profit column but visible in trade history and funding graphs. That's the gap the farming check covers, and a strong finding vetoes a wallet's Copy Score outright no matter how good the headline looks.

The honest summary: evasion is an arms race between sharps and bots, and the copier is the one paying for ammunition on both sides. Vet accordingly — the full checklist is here.

CopyGrade is analysis-only — it never executes trades or holds funds. Not financial advice.