What Is Copy Trading — and What Does It Actually Copy?

Copy trading means letting a second account mirror trades taken by another trader. In prediction markets, that usually means your wallet buys the same outcome tokens, in the same market, at roughly the same time as the trader you follow. In an on-chain model you keep custody of your money; in a platform model the platform holds it. Either way, the copied trader does not receive it and cannot withdraw it.

The copy is never exact. Your fill price, size, fees and exit can all differ, and a binary market can still go to zero. Copy trading transfers a stream of decisions, not a track record, and it does not transfer the leader's bankroll, information or tolerance for loss.

What is actually being copied

A copy tool watches the leader's fills and then places its own order. On Polymarket, that is a separate layer: the tool reads public wallet activity from the Data API and submits orders to the central limit order book. The leader's trade is public, but your order is a new order with its own price and queue position.

Different tools copy different things:

• New trades only, or new trades plus the leader's existing open positions.
• A proportional slice of the leader's size, or a fixed dollar amount per trade.
• The leader's exits, or only entries, leaving you to decide when to sell.
• Every market the leader touches, or a filtered subset.

The distinction matters because a leader's edge may live in position management rather than entries. If the tool copies the buy but not the sell, you are running a different strategy.

Regulators use three related terms. The International Organization of Securities Commissions describes copy trading as a follower choosing trades or traders to replicate, mirror trading as a more automated model that copies all trades of a chosen leader, and social trading as an information-sharing layer where the investor still decides whether to act. The terms are often used interchangeably, so read what the tool actually does rather than what it is called.

Two models: social copy trading and on-chain mirroring

In a platform or social copy-trading product, the broker or app holds leader and follower accounts and calculates the follower's size. Some broker models outside prediction markets can fill followers at the leader's price because the exposure is synthesized, for example through contracts for difference. A prediction-market copy tool cannot synthesize the outcome token; it has to buy it. You are relying on the platform's execution, solvency and leader-selection standards. You are also relying on its definition of the leader's return.

In on-chain mirroring, your wallet holds the outcome tokens and the tool submits transactions on your behalf. You can verify the leader's history directly, and you can stop at any time. But there is no netting or internal matching. Every follower who copies the same signal competes for the same order book, and the leader's own fill may already have moved the price. IOSCO calls this timing and pricing risk: when a leader has a sizable following, the followers' momentum can push the price away from the leader's entry.

Prediction markets make that competition visible. Polymarket prices are probabilities between $0.00 and $1.00, and the displayed price is the midpoint of the best bid and ask. If the spread is wider than $0.10, the last traded price is shown instead. You pay the ask when buying and receive the bid when selling, so the price displayed on the market page is not the price you get.

Wallet permissions and custody

Polymarket accounts use several wallet types. A Deposit Wallet is the default smart wallet for account wallets deployed on or after May 4, 2026. Legacy accounts may use a Proxy Wallet created through Magic Link or Google, or a Safe Wallet created with an external signer such as MetaMask or Rabby.

For copy trading, the important question is what the tool is allowed to do. A Polymarket Session Key is a separate signer that a Deposit Wallet owner can authorize for scoped, time-limited trading. It can be limited to the central limit order book, to Combos, or to all supported venues, and it cannot withdraw funds from the Deposit Wallet. Session Keys currently work only with Deposit Wallets and are in beta. That is a much narrower permission than handing over a seed phrase.

A useful rule: a copy service needs permission to trade, not permission to withdraw. If it asks for your seed phrase, your private key, or an approval to an unknown contract that can move your tokens, it has more power than the job requires. Approvals are permissions, and a malicious or compromised contract can use them.

Non-custodial copying also means the positions are yours. Stopping the copy does not close them. You still own the outcome tokens and still bear the resolution risk.

Where the risk actually comes from

The first risk is the market itself. A binary position bought at $0.40 pays $1.00 if the event happens and $0.00 if it does not. The maximum loss is the stake, and the maximum gain is the difference between the stake and the payout. There is no stop-loss that guarantees an exit at a chosen price in a thin book.

The second is the leader. A public track record can be luck, and a leaderboard can flatter a strategy that took large risks during a favorable window. IOSCO warns that leaderboards may mislead investors about a leader's expertise, the strategy used, risk-adjusted returns, past returns and the incentives the leader receives. In a controlled experiment by Jose Apesteguia, Jörg Oechssler and Simon Weidenholzer, 35 percent of participants chose to copy another participant. Of those copiers, 88 percent copied someone who had selected the riskiest asset, and the authors concluded that copy trading leads to excessive risk taking.

The third is execution. Your order can arrive late, fill partially, or walk several price levels. The fourth is strategy drift: IOSCO notes that leaders may change strategy or products without telling followers. The fifth is cost drag from frequent trading and taker fees. The sixth is operational and resolution risk, including wallet bugs, key compromise and disputed market outcomes.

A concrete comparison

Suppose a leader buys 10,000 Yes shares at $0.40, spending $4,000. If Yes resolves true, the position pays $10,000, a profit of $6,000, or 150 percent of the stake.

You want to commit about the same portfolio weight. With a $10,000 bankroll, a 4.2 percent position is $420. At $0.42, that buys 1,000 shares. If Yes wins, you receive $1,000 and make $580, or 138 percent. If No wins, you lose $420, or 100 percent of the position.

The 12-point gap between 150 percent and 138 percent is entry price alone. Add the taker fee and the gap widens. For a Politics market, the fee formula is shares multiplied by the fee rate multiplied by price multiplied by one minus price. At a 0.04 rate, 1,000 shares at $0.42 cost about $9.74 in fees, reducing the winning profit to about $570, or 136 percent.

The leader risked $4,000; you risked $420. That difference is deliberate. Copying the direction does not require copying the balance sheet.

What copy trading can and cannot do

It can save research time, enforce a consistent process, give a small account access to strategies it could not easily run alone, and diversify across more markets than one person can monitor. It cannot transfer the leader's skill, guarantee the same fill, make an illiquid market liquid, remove binary loss, or eliminate resolution risk.

The practical test is simple. If you cannot explain in one paragraph why the leader buys, how the leader sizes positions, and what would make you stop copying, you do not have enough information to copy yet.