How Polymarket works: order books, resolution, and fees for copy traders
Polymarket is a prediction market where outcome shares trade on an order book and pay $1 if the outcome happens, $0 if it doesn't. Three pieces of machinery sit under every trade — a hybrid order book, tokenised outcome shares, and an optimistic oracle that resolves markets — and each one changes what a copier actually receives when they mirror someone else's trades. This is the reference for how the machine works and where it bites.
How does trading on Polymarket actually work?
Polymarket runs a hybrid central limit order book (CLOB): orders are created and matched off-chain by Polymarket's operator, then settled non-custodially on-chain on Polygon, with positions collateralised in USDC. You sign orders from your own wallet; the operator matches them; the chain settles them. The practical consequences:
- Every fill is public. Settlement on-chain means any wallet's complete trade history is queryable — which is what makes copy trading, wallet vetting, and forensic farming detection possible at all.
- Liquidity is whatever the book holds. A price you see is only available for the size resting at it. The leader's fill price was available for the leader's size, at the leader's moment — a copier arriving later, taking liquidity, walks the book. That's slippage, and it always runs against you.
- Makers and takers are different roles. A patient trader resting limit orders (a maker) and a copy-bot crossing the spread to chase fills (a taker) experience different costs in the same market. Copy-bots are structurally takers.
What are outcome shares? (the Conditional Token Framework)
Outcome shares are ERC-1155 tokens minted through the Conditional Token Framework (CTF). The two primitives worth knowing:
- Split: $1 of USDC can always be split into one YES share plus one NO share of the same market.
- Merge: one YES plus one NO can always be merged back into $1 of USDC — at any time, with no trade on the book.
YES and NO prices therefore always sum to roughly $1, and a share's price reads as the market's implied probability (a YES at 62¢ ≈ 62% implied).
The merge primitive matters enormously to copiers, because merging is an exit that never appears as a sale. A trader holding YES can buy the matching NO, merge both back into USDC, and walk away — and a copy-bot watching for sells sees nothing. Polymarket's own newsletter describes sharp traders using exactly this to shake copiers, and it's one of the farming signatures we detect when used systematically. If your bot only mirrors visible sells, this is the blind spot.
Multi-outcome events (elections with many candidates, say) use a related mechanism — negative risk — with conversion rules of its own; the detail matters to market makers more than copiers, but it's the same idea: positions can transform without an on-book trade.
How do markets resolve? (UMA's optimistic oracle)
Polymarket markets resolve through UMA's optimistic oracle. After the real-world outcome, anyone can propose a resolution by posting a bond; if nobody disputes it within the challenge window, it stands. A dispute escalates to a vote of UMA token holders. For a copier this means:
- Resolution is not instant. Capital stays locked between an event finishing and the market resolving — proposal plus challenge window at minimum, longer if disputed. A strategy that looks capital-efficient at the leader's bankroll size can be a queue of locked positions at yours.
- Resolution risk is real but rare. Ambiguously-worded markets occasionally resolve against the apparent outcome, and disputes add delay. A wallet whose edge concentrates in fuzzy, contestable markets carries a risk its PnL doesn't show.
What does it cost to trade?
Costs come in layers, and the layers fall differently on a leader and a copier:
| Cost | Who pays it | Notes |
|---|---|---|
| Platform trading fees | Both, where applied | Polymarket's fee schedule has varied by market type over time — check the current schedule rather than assuming zero |
| Spread + slippage | Mostly the copier | Takers cross the spread; size walks the book |
| Latency drift | Only the copier | The price moves toward the leader's direction before your fill lands |
| Bot fees | Only the copier | Execution bots typically charge per trade — see choosing a bot |
| Gas | Effectively neither | Abstracted on Polygon; not a practical factor |
The headline return a leaderboard shows is computed on the leader's own fills with none of the copier-side layers — which is why a copier typically keeps 20–40% less than the headline, and why the Copy Simulator replays a wallet's actual fills with modelled fees, slippage, and your chosen latency instead of trusting the posted number.
Why mechanics are where copy trading is won or lost
Every mechanic above is a gap between the leader's result and your result: the book makes you pay spread the leader may have earned, latency hands back part of every move, merges let exits happen where your bot can't see them, and resolution locks your capital on someone else's schedule. Vetting a wallet — the full checklist — is largely the act of asking which of these gaps the wallet's edge survives. The Copy Score scores the post-gap number, not the headline.
CopyGrade is analysis-only — it never executes trades or holds funds. Not financial advice.