Anatomy of a farmed wallet: the signatures that turn you into exit liquidity

Farming is when a trader builds a position quietly, advertises a clean-looking record to attract copiers, and then exits into the demand those copiers create — turning them into exit liquidity. It is the single biggest risk in Polymarket copy-trading, and it leaves specific, on-chain fingerprints. Here are the ones that give it away.

Iceberg accumulation

The farmer builds a position in slices small enough that no single trade looks worth copying. By the time a visible, copy-tagged trade prints, they're already long — and your copy-bot is buying the position they're about to sell.

The pattern: a cluster of small sub-$200 buys on one market inside a short window, sized to sit under copy-bot alert thresholds. CopyGrade flags at least eight such buys clustered within 30 minutes.

Self-trade wash

The same operator takes both sides of a thin market. No net position changes hands, but volume prints and the wallet looks active and liquid — propping up a headline that isn't backed by real edge.

The pattern: repeated near-price buy/sell pairs on the same market and outcome inside a tight window — at least eight pairs within 10 minutes, well past ordinary back-and-forth trading.

Manufactured-volume churn

A wallet that looks busy because it is busy — but the activity produces no edge. Volume for the leaderboard, not for the copier. Copy it and you pay the spread on hundreds of trades that net nothing.

The pattern: a high trade count over the window — at least 350 trades — producing less than 1.5% net edge relative to the capital deployed.

Pump-and-dump

A position accumulated quietly, then exited into a price run-up the wallet's own flow helped drive. The copier mirrors the entry and becomes the exit. This one is weighted as suggestive rather than damning — it can genuinely look like skilled timing — so it nudges the risk level rather than hard-vetoing on its own.

The pattern: a buy cluster of $500 or more with buy prices walking up, then sells at a volume-weighted price 30% or more above the buy VWAP, the whole cycle inside 12 hours.

Copy-bait

The inverse of an iceberg: build quiet, then advertise loud. A conspicuous, mirror-me entry sized specifically to be copied, followed by a quick exit once followers pile in.

The pattern: several quiet sub-$200 buys, then a single headline entry of $2,000 or more, then an exit within six hours on the same market.

Decoy wallet clusters

A primary wallet and a handful of linked sub-wallets trade opposite sides of the same markets. The cluster nets near-zero edge overall, but the primary shows a clean record because the losses live on the decoys. You copy the winner; the operator keeps the whole cluster's flat result.

The pattern: a primary plus up to five sub-wallets that share at least three markets, co-trade within five minutes, take opposite sides, and net near-zero aggregate edge — caught by a cross-wallet pass and folded into the primary's risk level.

Cross-wallet coordination

Several wallets trading the same markets in tight time-lockstep without the clean decoy structure — pump rings and wash networks that move price together, then distribute the exit to followers. Also weighted as suggestive, because correlation isn't proof.

The pattern: three or more wallets co-trading several of the same markets within five minutes of each other.

How a finding caps the score

Not every flag is equal. Each is weighted by severity (how bad it is if real) and confidence (how far past its threshold the pattern sits), and the weighted total maps to a farming level. A moderate finding caps the Copy Score at 64 — the wallet can never read as a strong copy candidate. A high-confidence signature, or several moderate ones, caps it at 30, in the Avoid band, however good the headline PnL. One unmistakable signature can veto on its own; two borderline ones may only warn.

What this can't catch yet

Honesty matters more than coverage here. The thresholds above are heuristic priors, tuned against synthetic patterns and a small set of real outcomes, and they will be recalibrated. False positives are possible — a genuine scalper running tight spreads can trip the wash threshold without intent — which is why every finding cites the trades it was built from, so you can audit it yourself. And the strongest decoy signal, shared-funder linkage, needs an on-chain funding-graph source we don't yet ingest; until it lands, a loosely-coordinated cluster can still slip through. The full taxonomy, with each detector's exact rule, lives on the farming detection page.

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