Slippage and Latency in Copy Trading

A copied trade is a new trade, not a clone. Your fill differs because you act after the leader, your order consumes a different book, and your size may not fit. In thin prediction markets, the gap between the leader's price and yours can be larger than the entire edge you are trying to copy.

Slippage is not a small technical detail you can ignore while judging a track record. It is the price of admission, and it determines whether a copied strategy is profitable in your account rather than in the leader's.

The chain from the leader's fill to yours

The leader's order matches first. The trade is printed, and your copy tool has to notice it. That notification may come from a WebSocket feed, an API poll, or a platform event. Then the tool applies its risk checks, calculates your size, builds an order, signs it and submits it to the exchange. The exchange matches it against resting orders.

Every step takes time. On Polymarket, matching happens offchain and settlement happens onchain, and a marketable order can carry a matching delay before it trades. All orders are technically limit orders; a market order is simply a limit order priced to execute immediately against resting orders. That means your "instant" copy is still interacting with a specific book at a specific moment.

If the leader's order was large enough to move the price, the move has already happened before your order arrives. IOSCO describes this as timing and pricing risk: when a leader has a sizable following, the followers' momentum can push the price away from the leader's entry, so some copy traders buy higher or sell lower.

Order-book depth, with numbers

Suppose the asks for a Yes outcome are:

• 500 shares at $0.40
• 800 shares at $0.41
• 1,000 shares at $0.43

You want 1,500 shares. You take 500 at $0.40, 800 at $0.41 and 200 at $0.43. The cost is $200 plus $328 plus $86, or $614. Your average price is about $0.4093, not $0.40.

Now add the taker fee. For a Politics market, the fee is shares multiplied by the 0.04 rate multiplied by price multiplied by one minus price. For 1,500 shares at the average price, that is roughly $14.51. Your all-in cost is about $628.51, or $0.4190 per share.

Suppose the leader bought at $0.40 and you believe fair value is $0.43. At $0.4190, your expected profit is about 1.1 cents per share, or $16.50 on the position. The 3-cent edge has shrunk by almost two-thirds before you have sold anything.

Double the order to 2,500 shares. After taking the 1,000 shares at $0.43, the next 200 come from a higher level, say $0.46. The cost becomes $1,050 and the fee about $24.36, for an average of roughly $0.4297. At a $0.43 fair value, the expected profit is about $0.65 on 2,500 shares. Doubling the size turned a profitable copy into a rounding error.

That is the core lesson: the same signal is a different trade at different sizes. A leader's edge may exist only at the leader's size.

Latency and partial fills

Latency: where it bites and where it does not

Latency matters most when the market is repricing quickly. In-play sports markets, short-duration crypto up-or-down markets, breaking news and markets close to resolution can move several cents in the time it takes to detect and submit an order. If the leader bought at $0.50 and the price is $0.55 when your order arrives, the edge may already be gone.

Latency matters less in long-horizon markets. A political or economic market resolving in three months does not usually reprice in two seconds unless a genuine news event lands. Copying a thesis-driven position minutes later can produce a similar result. The dangerous combination is a short horizon, a thin book and a leader with a public following.

There is also a slower form of latency: your own decision to approve the copy. If you review each trade manually, you add minutes or hours, not milliseconds. That can be appropriate for long-horizon markets and disastrous for fast ones.

Partial fills and adverse selection

A limit order can fill partially, and the copies you get may not be a random sample. Suppose you place a limit at the leader's $0.40. If the price falls to $0.38, your order fills and you are immediately down 2 cents. If the price rises to $0.42, your order does not fill and you miss the winner.

This is adverse selection. The copies you complete are disproportionately the ones where the market moved against the position. A market order avoids non-fills but pays the impact instead. Neither method gives you the leader's fill.

Partial fills create a second problem: you may end up with a smaller position than intended, or a position built from several prices. If the tool does not reconcile the actual fill, your risk budget can be wrong in either direction. Always size from the filled amount, not the requested amount.

Markets where copying is structurally hard

Some markets are poor candidates for copying regardless of the leader's skill.

• Wide spreads. Polymarket displays the midpoint of the best bid and ask, but if the spread is wider than $0.10 it shows the last traded price instead. A wide spread means your entry and exit cross a large gap that the leader may not have paid.
• Thin books. If your order is large relative to visible depth, your own trade moves the price. The leader may have been the only real liquidity.
• Very short duration. An up-or-down market that resolves in an hour is a different bet by the time a follower reacts. The resolution window, not just the price, has changed.
• Prices near $0.01 or $0.99. At $0.02, a one-cent tick is a 50 percent move in the price. Fees and ticks dominate the economics.
• Ambiguous resolution. Polymarket uses UMA for prediction markets, with a two-hour challenge period and a disputed-resolution path that can take four to six days. A market can also resolve 50/50 in rare cases. A leader who plans to hold through resolution and a follower who needs liquidity earlier are not running the same trade.
• Market-making leaders. If the leader is providing liquidity on both sides, each fill may be inventory management rather than a directional signal. Copying every fill can mean paying the spread twice.
• Multi-outcome or hedged positions. Copying one leg of a hedge without the other leg changes the payoff completely.

Measure, then reduce the gap

Measure your own implementation shortfall

For every copied trade, record the leader's price, your average fill, the fees, the delay and the filled fraction. The implementation shortfall is the difference between your average price and the leader's, multiplied by shares, plus fees, plus the opportunity cost of the part you did not fill.

If the median shortfall is larger than the expected edge per share, the strategy is not copyable at your size. That is a more useful conclusion than another month of hoping.

Track the shortfall by market category, time of day and order size. You may find that copying works in long-horizon politics but fails in in-play sports, or that it works below a certain order size and breaks above it. Those are actionable boundaries.

How to reduce the gap

Use limit orders with a maximum price deviation from the leader's fill. If the market has moved beyond your limit, let the copy fail. Size from visible depth, not from the leader's notional. Split large orders rather than sweeping several levels. Avoid the first and last minutes of a market, when spreads are widest. Prefer markets where the thesis takes days or weeks to play out. Accept that some trades are simply uncopyable, and do not force them.

The best copy tools will show you the leader's price and your fill side by side. If yours does not, build that record yourself. You cannot manage a gap you cannot see.