Risk Controls for Copy Trading

Risk controls are the rules you set before a trade that cap what one market, one leader or one bad week can do. The essential set is small: a maximum loss per position, a maximum exposure per market and per event, a drawdown stop, a kill switch, and written pause rules.

The purpose is not to avoid losses. Losses are part of the strategy. The purpose is to make sure no single loss, and no cluster of correlated losses, removes your ability to keep operating.

Set limits in maximum loss, not notional

In a binary prediction market, the maximum loss on a long position is the stake. A Yes position bought at $0.40 can go to zero, and the most you can lose is what you paid. That makes the risk budget easy to express: decide the dollars you are willing to lose, then buy the number of shares that costs that amount. The CFTC's investor guidance is blunt: trade only with risk capital.

A workable starting framework for a $10,000 bankroll is 1 to 2 percent maximum loss per position, 2 to 5 percent per market, 5 to 10 percent per event, and 25 to 50 percent total at risk. At the low end, that is $100 to $200 per position, $500 per event and $2,500 across the portfolio.

Check the limits before the order, not after. Add the new position's worst-case loss to the existing positions that would lose in the same scenario. If the sum breaches a cap, reduce the order or skip it. Do not count unrealized gains as risk capacity; a mark-to-market gain is not cash until sold or resolved.

Maximum market and event exposure

A market is one question with one resolution. An event is a group of related markets that share a common cause. The distinction is where most concentration hides.

Consider an election. You might hold a market on who wins the nomination, a market on which party wins the presidency, a market on a particular state, and a market on turnout. Four positions, four order books, but one underlying event. If each position risks $150, the cluster risks $600. If your event cap is $500, you are over the limit even though no single position looks large. Cut each to $125 or drop the weakest.

Ask what single outcome would make all of the positions lose, then sum the losses in that scenario.

Add a liquidity limit. If your position is large relative to the visible book at your exit price, the mark-to-market value is not what you can realize. A workable cap is 5 to 10 percent of the resting depth within a few cents of the current price. Polymarket's order book has no fixed size limit, but large orders can move the price significantly, and the documentation tells traders to check depth before trading in size.

Drawdown, kill switch and pause rules

The drawdown stop

Drawdown is the decline from a peak in account equity. If your account peaks at $10,000 and falls to $8,800, the drawdown is 12 percent. A drawdown stop says: at this point, stop adding risk.

A simple rule is a 10 to 15 percent peak-to-trough stop. At a $12,000 peak and a 10 percent stop, the trigger is $10,800. When equity touches it, you stop new copies, cancel resting orders and review. Do not move the peak down to avoid the trigger; that is the same as not having a stop.

The arithmetic of recovery explains why the stop matters. A 10 percent loss needs an 11.1 percent gain to get back to even. A 20 percent loss needs 25 percent. A 50 percent loss needs 100 percent. The deeper the hole, the more the strategy has to earn just to return to the starting point.

Distinguish a strategy drawdown from an operational failure. A series of losing resolutions is variance, and it may be within the historical range. Bad fills, stale prices, missing data or a leader who has changed strategy are process failures, and they justify stopping immediately regardless of the profit and loss.

Have a written reset rule. After a stop, resume only after a review, a cooling-off period and, ideally, a period of paper copying at half size. The reset should be a decision, not a reflex.

The kill switch

A kill switch is one action that stops new orders and cancels open orders. In a stronger form, it also closes positions and revokes the copy tool's trading permission.

It must be independent of the copy logic; otherwise a bug can disable both. Test the switch on a schedule, and make sure it works when the user interface or API is degraded. On Polymarket, a Session Key can be scoped and time-limited and cannot withdraw funds; revoking it is a clean way to remove trading permission without moving the money. Token approvals are a separate permission and should be reviewed too.

Be honest about what a kill switch cannot do. It cannot guarantee an exit price in a thin book, it cannot close a position that has already resolved, and it cannot remove resolution risk. It stops new risk; it does not erase existing risk.

Define two levels. A pause stops new orders but keeps positions open, which is appropriate when you want to think. A kill flattens or attempts to flatten everything, which is appropriate when the process is compromised. Write down which one you will use and who can trigger it.

Pause rules

Automatic pauses should trigger on process signals, not on a fixed number of losses. A losing streak is often just variance, and pausing after three losses can systematically remove you from a strategy right before it recovers.

Useful automatic triggers include:

• The leader changes market category, holding period or typical position size.
• Your average slippage exceeds a set threshold over the last ten copies.
• Your fill rate falls below a set level, or fills are consistently worse than the leader's.
• Price or position data goes stale for more than a defined interval.
• A market enters its final day before resolution, or a resolution dispute is opened.
• Platform volatility, an API outage or a matching-engine restart.
• The portfolio hits its drawdown or daily loss limit.

Manual triggers include a thesis that no longer holds, a news event that changes the market's basis, a leader whose last few trades you cannot explain, and anything that makes you unsure whether the data you are seeing is current.

IOSCO's review of copy trading identifies unannounced strategy changes, misleading leaderboards, conflicts of interest and operational risk as potential sources of investor harm. A pause rule that forces you to re-read the strategy after a change is a direct response to the first of those.

What can and cannot be automated

Software can aggregate exposure across positions, enforce maximum-loss caps, reject orders outside a price limit, cancel and replace orders, calculate drawdown, revoke a session key and send alerts. It can reconcile expected fills against actual fills and log every decision.

Software cannot judge whether a resolution is ambiguous, whether a leader has changed intent, whether a news event invalidates the thesis, or whether you personally can tolerate the drawdown. It also cannot eliminate its own failure modes: stale marks, API outages, partial fills, race conditions between cancel and fill, and oracle disputes.

Automate the arithmetic and the permissions. Keep judgment and the kill-switch authority with a person.

A control sheet

1. Maximum loss per position: 1 to 2 percent of bankroll.
2. Maximum per market: 2 to 5 percent.
3. Maximum per event cluster: 5 to 10 percent.
4. Total portfolio at risk: 25 to 50 percent.
5. Drawdown stop: 10 to 15 percent from peak, with a written reset rule.
6. Daily loss limit: 3 to 5 percent.
7. Slippage and fill-rate limits for copied trades.
8. Liquidity cap based on visible depth.
9. Kill switch tested monthly, with a defined pause level and kill level.
10. Session key scoped to trading, with an expiry and a revocation procedure.
11. Daily reconciliation of positions, fills and fees.
12. Written pause and resume rules.

Controls do not make a bad strategy good. They make a good strategy survivable and a bad one small enough to learn from.