Copy trade safely

Copying a wallet is not a strategy. Copying a wallet whose edge you have checked, at a size you can actually fill, on markets slow enough that your fill resembles theirs, is a strategy. This page is the method — five steps, each one a thing you can check before you risk anything.

We sell the data and the method. We do not sell picks, we do not promise returns, and nothing on this page is advice about what to trade.

1. Qualify — get to a shortlist that is not survivorship

Start on the screener, and start with a bar, not a ranking. Anybody can find an impressive wallet by sorting a leaderboard; the point of the first step is to find the wallets that would still look impressive if you had picked them a year ago.

A workable first screen:

• min_resolved_bets at 100 or more. Below that, nothing downstream means much.
• max_skill_qvalue at 0.05 or 0.1. This is the false-discovery-rate control — it is what stops you from picking the wallet that got lucky out of three million tries.
• min_ev_ci_low above 0. The lower bound of expected value per dollar being positive is the honest test of an edge: the point estimate can be flattered by a sample, the lower bound cannot.
• trader_kind=human unless you specifically want machines. A market maker's entries are not a directional opinion, and a bot's cadence may be impossible to follow.
• max_avg_bet_size at something you could fill. A wallet averaging $240,000 a position is not copyable in a thin book no matter how good it is.

The shipped Copyable humans screen is exactly this shape and is a reasonable starting point to modify rather than a recommendation to follow.

2. Verify — open the dossier and try to talk yourself out of it

A screener row is a hypothesis. The dossier is where you attack it. Five things to read, in this order:

The verdict chip

Every wallet carries one classification — human, bot, market maker or unclassified — with the counts that produced it attached, so you can re-litigate the label instead of trusting it. A bot is at least 10,000 lifetime fills across 250 or more markets; a market maker earns maker rebates at least five times its taker rebates at machine scale.

This matters before anything else because it changes what the numbers mean. A market maker's PnL is a spread business. Copying its entries copies the wrong half of its strategy.

The risk ratios

Sharpe, Sortino, Calmar, max drawdown and profit factor, computed over the wallet's daily realized-PnL curve. Two caveats are printed on the panel and worth repeating:

• They are PnL-based, not return-on-capital. We do not know the capital behind a wallet's positions, so these ratios describe the shape of its profit curve, not its return on equity. They compare wallets to each other; they do not compare a wallet to your brokerage account.
• They refuse to compute on a thin sample. A ratio needs 20 active days, and an annualized one needs 90 days of span. Below that the panel says what is missing rather than printing a number built on four data points.

Max drawdown is the number to sit with. It is the answer to "how bad did it get, in dollars, before it got better" — and if you would have quit at that number, you were never going to capture this wallet's edge.

Best and worst

The calendar view surfaces the wallet's best and worst days. Read the worst one first. A wallet whose entire edge is two enormous days is a wallet whose edge you will not be present for.

Timing

The timing histogram shows when in a market's life the wallet enters. Early entries on long-horizon markets are copyable — you have hours or days to take the same side at a similar price. Entries in the last minutes of a short market are not: by the time the fill is indexed and you have read it, the price that made it a good trade is gone.

The cluster

The network view shows funding relationships. Two things to look for: whether your candidate is one of several wallets funded by the same source (in which case its "independent" record may be one operator's record counted several times), and whether the wallets you are about to copy are actually the same wallet wearing different addresses.

3. Size — on anything except the win rate

Sizing is where most copy trading dies, and it dies for arithmetic reasons rather than analytical ones.

• Copy direction, never dollars. A wallet putting $500,000 on an outcome is telling you its probability estimate differs from the market's. It is not telling you your position should be $500,000, or any particular fraction of it.
• Size against the drawdown you read, not the profit. If the wallet's worst peak-to-trough was 40% of its cumulative PnL, assume yours will be worse, because you will be entering later and exiting sooner.
• Assume you capture a fraction of the measured edge. Between the fill you will not get and the timing you cannot match, the sensible planning assumption is that a meaningful part of the historical edge does not survive to you. Size as if that is true and you will still be trading in three months.
• Check the book before you commit to a size. The ladder on the market page shows cumulative depth and previews impact at $100, $500 and $1,000. If your intended size walks the book, your entry is not the entry you are copying.

Want a number rather than a heuristic? Replay the signal with your own fee and slippage assumptions using Backtests, and look specifically at how much of the edge survives a slippage assumption you would defend out loud. That gap is the finding.

4. Watch — with a setting you will not mute

Follow the wallet, then decide what should reach your phone. Following alone delivers nothing; push is opt-in per channel.

For copy trading specifically, filter on size rather than score — a large position is a clearer statement of conviction than a threshold on a model output — and pair any win-rate floor with a sample floor. The Sharp money preset is that shape: position at least $5,000, from wallets over 60% across at least 20 resolved bets, capped at 15 a day with 30 minutes between.

If you are automating, take the webhook channel instead of polling, and verify the signature before acting on a payload.

5. Act — and record why

Execution is yours. Order entry on the market page is in beta and off by default (see Terminal guide); until it is on for your account, the market page hands you a link to execute on Polymarket, and the API executes nothing at all — it holds no custody and submits no orders.

Whichever route you take, write down the thesis before you click: which wallet, which check convinced you, what price you got against the price they got, and what would make you stop. After twenty copies, compare your realized PnL against theirs on the same trades. If you are capturing far less of the edge than you assumed in step 3, the problem is the method or the market's speed — change one of those rather than the wallet.

The failure modes, compactly

Next

• Screener — the ranking math behind step 1.
• Alerts — presets, per-wallet filters and the opt-in model.