When to stop copying a Polymarket wallet: the exit checklist
Stop copying a Polymarket wallet when any of five tripwires fires: its farming-risk check stops being clean, its CopyGrade Score decays out of the band you vetted, its drawdown breaches the stop you wrote down before you started, its realistic post-fee edge falls below the copy floor, or its behavior visibly changes — new categories, new sizing, a sudden burst of churn. The rest of this guide is how to define each tripwire precisely, how to monitor them without living in a dashboard, and how to exit without donating your gains to fees. It exists because the copy-trading loop everyone repeats — shortlist, vet, choose a bot, size, monitor — has a last step that almost nobody specifies, and the last step is where the money is lost slowly instead of quickly.
Write the exit before the entry
Every rule below works only if it's written down before money moves — afterwards, every breach comes with a story about why this time is different. This is the monitoring half of the contract you made when you sized the position from the drawdown: you sized for the wallet's worst historical stretch, so the stop is already implied. Make it explicit: "I exit if X," with numbers, in writing.
The five tripwires
1. The farming check stops being clean
This one is absolute. A wallet you copied because its farming-risk check was clean is a different wallet the day the check flags — the entire premise of the copy is gone, and the flag's failure mode (a record built to monetise its copiers) is the one that specifically targets you. The check is recomputed from scratch on every sync, so a new flag is current information, not history. Treat a severe flag as an immediate exit and a watch-level flag as an immediate size-down and review. (As always: a flag is our model's algorithmic risk assessment, not an accusation of conduct — but you don't need an accusation to stop copying someone.)
2. Score decay through a band
Don't react to single-point score noise — scores recompute hourly and wobble. React to band changes: the wallet you vetted as a copy candidate no longer grades as one. A band is the model's summary of everything at once, which makes it the right coarse tripwire; the public verdict page's band timeline shows you whether a dip is a blip or a slide.
3. Your drawdown stop, breached
Not the wallet's historical max drawdown — your stop, the one implied by your sizing. If you allocated so that a repeat of its worst stretch was survivable, then a new drawdown materially beyond that stretch means the historical record you vetted no longer describes the wallet. The arithmetic that makes this urgent: a 50% loss needs a 100% gain to recover. Exits on drawdown feel worst and matter most.
4. Edge below the floor
A realistic post-fee edge below about 1% means copying nets approximately nothing even when it works — that's why our scoring caps sub-floor wallets out of the candidate bands. If the wallet's realistic edge has slid under the floor and stayed there across multiple syncs, the copy has become a fee-generation machine: you pay taker fees both ways to capture an edge that rounds to zero.
5. Behavior change
Records are only predictive of the strategy that produced them. A baseball specialist suddenly heavy in crypto price markets, a $50-a-trade wallet suddenly sizing $5,000, a patient futures holder suddenly churning — each means the thing you vetted no longer exists, whatever the score says this hour. The wallet's category mix and trade cadence are on its public verdict page; a drift there is a re-vet trigger even when every number still looks fine.
Monitoring without living in a dashboard
Checking manually every day doesn't survive contact with real life, which is why the tripwires map onto watchlist alerts: the free tier includes farming-flag alerts (tripwire 1 — the one this guide calls absolute) on up to three watched wallets, while score-decay and drawdown alerts (tripwires 2 and 3) and Telegram/webhook delivery are Pro. Either way, an alert fires the moment a sync produces the change. Every analysis surface behind the tripwires is free regardless: the scores, verdicts, band timelines and farming forensics are public for every wallet. A reasonable cadence on top of (or instead of) alerts: a five-minute weekly review of each copied wallet's verdict page for tripwires 4 and 5, and a full re-vet monthly — the same checklist you used to enter.
Exiting cleanly
Three mechanics worth knowing before you hit the kill switch. Fees: your exit legs pay taker fees like every other fill — factor them in, but never let a fee delay a farming-flag exit; the fee is bounded, the flag's downside isn't. Open positions: stopping the bot stops new copies; decide explicitly whether to hold or unwind the positions already open, because "ride them to resolution" is a real choice with real risk, not a default. The post-mortem: rerun the wallet through the Copy Simulator over the period you copied it — comparing what the simulator says you should have made against what you actually made tells you whether your losses were the wallet's edge failing or your own latency and slippage, which determines whether the fix is a different wallet or a different bot.
Stopping is the system working
Across the whole graded universe, the Copyability Index read 1.5% at its 2026-08-15 launch (the page always carries the live value) — and membership in that small group churns. Most copies should eventually end; a clean, rule-driven exit isn't a failure of the strategy, it is the strategy. The alternative — holding through a farming flag because the first month went well — is how copiers become exit liquidity.
A CopyGrade Score, verdict, and farming flag are documented research opinions computed from public Polymarket data, recomputed on each sync — our own validation of whether the score predicts copier returns is still null. CopyGrade is analysis-only and never executes trades or holds funds. Not financial advice.