Copy Baskets
Copy a diversified portfolio of leaders with allocation weights, an aggregate stop-loss and a pre-creation correlation check.
A copy basket lets you copy a *portfolio* of leaders instead of betting your copy allocation on a single trader. This doc explains how baskets work, the risk controls built into them, and why we built them at all.
## Why Baskets Exist
The uncomfortable statistic first: roughly **80% of Polymarket participants lose money**. Any single leader you copy β however good their last month looks β sits inside that distribution, and single-leader copy is fragile: one cold streak, one category they don't understand, one abandoned account, and your whole copy allocation goes down with them.
A basket spreads that risk across several leaders with explicit weights and an aggregate loss limit. It doesn't make copying safe β nothing does β but it turns "I'm betting on one stranger" into "I'm betting on a weighted panel, with a circuit breaker".
## What a Basket Is
- **2 to 8 leaders**, each with an **allocation weight** in percent (weights sum to 100%).
- β’A **per-trade cap for the basket**. Each member's effective cap is the basket cap Γ their weight β a leader with 30% weight in a $50-cap basket can trigger at most $15 per copied trade.
- β’An **aggregate stop-loss in USD**. When the basket's combined realized losses hit this number, the entire basket **auto-pauses** and you're notified on every channel (bell, Telegram, push).
## The Diversification Check
The basket builder has a **"Check diversification"** button. It computes the pairwise correlation of the leaders' daily on-chain PnL and warns you about any pair above **0.70** β in plain terms: "these two are the same bet twice".
This matters because the failure mode of naive baskets is picking five leaders who all trade the same election markets in the same direction. That's not diversification; it's one position with five names on it. The check uses on-chain history, so it reflects what leaders actually did, not what they claim.
Honest limit: correlation is computed on daily realized PnL and is backward-looking. Two leaders can be uncorrelated historically and still pile into the same market tomorrow. Treat the check as a filter for obvious duplication, not a guarantee of independence.
## Per-Leader Controls Still Apply
Inside or outside a basket, each leader you copy has its own controls:
- **Copy mode**: proportional (X% of the leader's trade size) or fixed ($Y per trade, regardless of the leader's size).
- β’**Per-leader stop-loss** β a USD loss limit on that leader alone.
- β’**Max slippage %** β if the current price has moved too far from the price the whale got, the copy is skipped. When live CLOB book data is available, the comparison uses the live book price; this is what protects you from copying a whale move that's already 10 minutes stale and fully priced in.
## Pausing and Resuming
When the aggregate stop-loss fires, the basket pauses β no new copies from any member. **Resuming is a deliberate human action**: you go to the basket, read the damage, and explicitly resume it. There is no auto-resume, and the resume is written to the audit log. The point of a circuit breaker is the pause *and the moment of reflection*; software that quietly turns itself back on defeats both.
## Plan Limits
On the Bot plan you can run up to **3 baskets** and **10 standalone leaders**. Baskets and standalone follows are counted separately.
## What Baskets Don't Do
- They don't vet leaders for you. Weighting three bad traders still loses money, just more smoothly.
- β’They don't rebalance dynamically β weights are yours to adjust.
- β’They don't bypass the platform's automation caps: every copied trade still passes the same per-trade, per-day, and global limits as any bot trade (see [Live CLOB Trading](/docs/live-trading)).
## Related Docs
- [Copy Trading](/docs/copy-trading)
- β’[Whale Tracker](/docs/whale-tracker)
- β’[Trader Profiles](/docs/trader-profiles)
- β’[Live CLOB Trading](/docs/live-trading)