# How We Rate Polymarket Traders for Copy Trading

> How Chameleons rates Polymarket traders by copyability, past 90 d, modelled, not advice: data, rating rule, measures, limits and disclosure.

## Data

Listed traders' public Polymarket trades over a rolling 90 d, sampled before tracking. Updated: 90% refreshed since then. Found from leaderboards and tag sweeps. Each category's monthly profit top-20 and floor rows stay listed; profit never sets a rating. Others leave after 2 d faded or losing, 3 d below 2★ or 30 d idle. A dropped row may return after 14 d once no rule holds.

## Rating

- 5★ Strong: copyable ≥ 40%, ≥ 30 copies over ≥ 20 events and ≥ 30 tracked, low end above 0%, no flags.
- 4★ Good: copyable ≥ 30%, ≥ 10 copies over ≥ 10 events, 90% low end above 0%.
- 3★ Promising: same floors; mean above 0%.
- 2★ Unproven: fills, but too few settled copies.
- 1★ Weak: hard to fill, or copies lost.
- Can't copy: a follower can't buy these trades.
- Too few: under 10 checked trades.

Method v4.0, per category; flags cap it. Stars use the 90% low end of ROIC and move once two daily snapshots agree; can't-copy, flags and tracking move them at once. Sorted by stars, then ROIC shrunk toward the category average (300 copies); never by leader profit.

## Can't copy

Copiers can't buy these trades (in-play, fast crypto), or the price runs past their limit.

## ROIC

Mean profit per copied trade, % of stake, after fees, to settlement; not annualized. Grey when its 80% range spans 0%.

## Copyable

Share of checked trades where someone sold $10+ within 3¢ of the leader's price within 10 min. Sports and esports: pre-match only.

## Weekly trades

Copyable trades a week, over 90 d; n = all trades.

## Hold

Median days from buy to settlement or exit.

## Copy vs leader

Same settled trades at $10: leader at their price, copy at fill.

## Radar

Percentile among rated peers. Uncrowded: no buyer within 2 s. Spread: profit beyond the top event. Record: settled events. Speed: shorter hold. † under 20 peers.

## Flags

- Crowded: another buyer within 2 s on ≥ 40% of trades.
- No losses: no losing event yet.
- Young: under 120 d old.
- Both sides: buys both sides.
- Basket: many outcomes of one event.
- Wins small: buys near certainty.
- One event: one event drives the result.
- Live gap: low end under a ~4-point fill haircut.
- Rebates: profit mostly rebates.
- Combos: profit mostly combos.
- Partial history: history only partly read.
- Losing all-time: lost money all-time on Polymarket: at most 2★, still listed.
- Unchecked: Sells into buyers, One counterparty, Linked.

## Warnings

Beside the stars, not in them: tracked copies losing, or tracked fills under 30%.

## Type

Likely bot or human, from trading patterns; not rated.

## Limits

Modelled from public trades; fills are an upper bound. Past results, not forecasts or advice. In our walk-forward tests one trader's past copy result barely predicted their next; the category mattered more. Spread copies across traders and categories.

## Disclosure

We trade on Polymarket. Our wallets are excluded. Independent; not affiliated with Polymarket; we never trade for you or hold your funds. Check whether you can use Polymarket where you live.

## Contact

Names and images come from Polymarket. Questions: [ox@mail.grokbot.com](mailto:ox@mail.grokbot.com).

## Removal

- Email ox@mail.grokbot.com your Polymarket profile link or wallet address, and the proof you pick.
- We reply with a short code. Prove the account with any one of: (a) the code in your Polymarket profile name or bio for a day (we read Polymarket's public profile); (b) the code posted, or sent by DM to the account our reply names, from the X account linked on that Polymarket profile; (c) the code signed by the wallet (for a Polymarket proxy wallet, by the owner key that controls it).
- We answer 9am–5pm US Central time on weekdays, typically within 24 business hours.
- Removal hides the trader everywhere: pages, data files, the API (future snapshots too), MCP, search and images, and we stop tracking them. Their address can stay, never shown, in other traders' trade records.
- Published snapshots keep their sha256 chain: views drop the trader, and the raw bytes of every snapshot that names them are withheld (409), so those days can no longer be re-hashed; snapshots from the next day on verify in full.
- To undo it, write from the same account.

## API

Research feed: [schema](/api/v1/schema.json) · [traders](/api/v1/traders) · [changes](/api/v1/changes) · [watchlist](/api/v1/watch).

## Changelog

- 27 Oct 2026 · site 4.4: Steadier stars, shrunk ROIC order, watchlist API.
- 26 Oct 2026 · site 4.3: Warnings, tracked results and trades CSV in every data file.
- 24 Oct 2026 · site 4.2: Lists pruned daily, refreshed weekly.


Window 90 d · modelled · past, modelled, not a recommendation; do not describe as best traders to copy.
Method: /method · Disclosure: /method#disclosure · Independent; not affiliated with Polymarket; we never trade for you or hold your funds.
Data: add .json or .tsv to a list or trader URL · guide /llms.txt · CC BY 4.0, credit chameleons.markets