How Shark Radar ranks skill (not luck)

Polymarket's native leaderboard (and most trackers) rank raw profit and volume. A wallet can top that board risking huge size at $0.97+ favorites — genuinely green, useless to learn from. We build a skill radar instead. Here's how it works.

Quotable definitions

ROI on risk: Return on the money a wallet actually risked (cost basis), not raw profit. Bankroll size stops buying the top of the board.
Luck filter: We strip records that look like one-off jackpots or incomplete samples — a single lucky season is not skill.
Uncopyable styles: Profit a follower cannot capture: market-making spreads, reward farming, split/merge arb, near-certain favorites at tiny edges, latency/HFT. Shows up as PnL; fails as something worth mirroring.
Archetype: A label for how the wallet actually trades (e.g. Sharp, Favourite buyer) so you can tell judgment from structural edge.
Confidence: How solid the record looks after the product gates. Public board indexability: Strong only.

The problem with every other board

Past profit barely predicts the next period: across 157 top wallets the correlation between past and future returns was 0.035, and copying the top-15 profit-ranked wallets into fresh markets returned −13.6% out of sample (sources on our copy-trading page).

What we score instead

Shark Radar scores every qualifying wallet on ROI on money actually risked, behind the product's statistical gates and an archetype classifier that labels style. Rankings recompute daily from on-chain Polymarket data.

We are not ranking "who made the most dollars." We are ranking "whose returns look like judgment on money risked — after luck and uncopyable styles are filtered out."

What gets someone on the board vs rejected

On the board: Clears the product gates; shows durable ROI-on-risk; style is labeled; confidence can be Strong or Moderate on the live table.

Rejected / filtered: Lucky streaks, incomplete records, and trading styles you couldn't copy anyway (MM, farming, structural arb, etc.).

Indexed public profiles (SEO): Strong confidence only — Moderate can appear on the live board UX but stays out of sitemap/index.

Exact numeric cutoffs live in product code — this page does not invent them.

Follow judgment, not fills

Blind copy fails on price: one profitable tennis wallet flipped from +$191k → −$64k when entries were just 2¢ worse — the fill a copier actually gets (copy-trading page). So we collapse fill-spam into net position changes and send one Telegram alert per real decision, not one alert per microscopic fill. Read-only. No keys. Not financial advice.

How to use it

Open the skill leaderboard.
Open a wallet profile — style, confidence, track record.
Watch traders you trust; get alerted when they take a large net position.
Research the market yourself at your price.

More context: how to track Polymarket wallets, vs Polyrank, vs Polycopy, and pricing.

Common questions

How is this different from the Polymarket leaderboard?

They rank profit and volume. We rank ROI on money risked after luck and uncopyable filters.

What is ROI on risk?

Return relative to capital actually put at risk — not headline PnL. Bankroll size stops buying the top of the board.

What is an uncopyable style?

Edge that disappears when you try to follow: market-making, reward farming, split/merge arb, near-certain favorites at tiny edges, latency/HFT.

Do you copy trades for me?

No. Alerts and analytics only — read-only, no keys, not financial advice.

How often do rankings update?

Daily recompute from on-chain Polymarket data.

Which wallets are publicly indexed?

Strong confidence profiles only (SEO policy). Moderate can still appear on the live board.

See which traders clear the skill bar

Strong-confidence profiles, luck-filtered, updated daily. Watch your first traders free.

Open the skill board