Methodology

01The data we start from

OrcaLayer indexes every Polymarket trade directly from the Polygon blockchain into our own database — over 1.2 billion+ trades and roughly 1.3 million+ marketsto date. We do not consume a third-party leaderboard or API; we read the chain itself, which means our numbers do not inherit anyone else's errors or omissions.

The indexer is split into focused jobs for cost and reliability: one polls Polygon blocks every few seconds for trade events (the source of truth for analytics), another tracks the CTF split / merge / redeem operations that move value in and out of multi-outcome markets. On top of the raw chain we layer Polymarket's own activity feed to fill any gaps, de-duplicated against the on-chain record so nothing is double-counted.

Before any trader is scored, four Polymarket protocol contracts — the CTF Exchange and NegRisk router contracts that sit on both sides of ordinary trades — are stripped out everywhere. Left in, a router can masquerade as a single hyper-active “insider” wallet; we learned this the hard way when one surfaced on a high-profile market, and now exclude them by default across the entire platform.

02Who we track: the 10-market gate

A wallet only enters our analytics once it has traded at least 10 distinct markets that have already resolved — markets with a known, settled outcome.

The reason is simple: luck. Someone who bets on a single market and wins shows a “100% win rate,” but that is noise, not skill. Requiring a minimum track record of settled markets filters out one-hit wonders and leaves traders whose results actually mean something. This gate is invisible to you — by the time a wallet appears on the Leaderboard, it has already cleared it.

03Smart Money: a three-part test

A wallet earns the Smart Money badge only when all three are true at once:

• Win rate of 55% or higher — measured per market, not per trade (see §05 for why that distinction matters).
• Positive total P&L — the wallet is actually up across its resolved markets, not just frequently right on small positions.
• Not classified as a farmer or bot — the Farmer Filter (next section) flags farmers, high-frequency bots, and router contracts, and none of them can earn the badge.

Note what this is not: it is not an opaque, weighted “AI score” you have to take on faith. It is a transparent rule — three conditions, all checkable against the same public data we show you.

04The Farmer Filter

This is the difference between OrcaLayer and a raw leaderboard. Standard rankings are dominated by wallets that look brilliant and are not. We detect and separate them with three independent mechanisms.

A deeper, dedicated write-up lives on our Farmer Filter page, with the full data story in our analysis of 470M trades.

05NegRisk-corrected win rate & P&L

Polymarket's multi-outcome markets (a single event with many possible winners, e.g. “Who wins the election?”) are built as a set of linked Yes/No outcomes under the hood, using a mechanism Polymarket calls NegRisk. Counted naively, a single position in one of these markets can be double-counted across its linked legs, inflating both win rate and P&L.

We correct for this in two places. The cash a trader moves when splitting or merging a multi-outcome position applies to both linked outcomes equally, so we divide it across them rather than counting it twice. And the resolved profit on a NegRisk market is halved so a single outcome is not credited as if it were two. The effect: a multi-outcome win counts as one win, exactly as a human would tally it — not the artificially boosted number a raw query produces.

Win rate itself is then the share of a wallet's resolved markets that ended in profit — markets, not individual fills. A trader who scaled into one winning market across 40 buys gets one win, not forty.

06FIFO profit & loss

Realized P&L is computed with first-in, first-outlot accounting — the same principle a tax accountant uses. For each wallet and market we walk every trade in time order: a buy opens a lot at its price, a sell closes the oldest open lots first, and a redemption at settlement closes whatever remains. The result is an honest realized profit or loss per position, an average entry price, and a clear winner / loser verdict — rather than a crude “value in minus value out” that ignores the order trades actually happened in.

07Median hold time

For each wallet we measure how long it typically holds a position — the time from its first trade in a market to its last — and report the medianacross all of that wallet's closed positions. We use the median rather than the average so one unusually long or short position cannot distort the figure, and we only show it for wallets with at least three closed positions, so it reflects a habit rather than a coincidence. It is a quick read on style: a few hours suggests a scalper, several weeks suggests a conviction holder.

08Matching Polymarket's own numbers

Wherever possible we reconcile our blockchain-derived results against Polymarket's own realized closed-position figuresfor a wallet, so the win rate and profit factor we display match what a trader sees on their own Polymarket profile. When we hold a near-complete copy of a wallet's closed positions, those reconciled numbers take precedence; when our copy is partial, we fall back to our own FIFO-derived figures rather than show something incomplete. Either way the source is transparent — there is no invented number in between.

09ISW Ukraine frontline signals

For Ukraine territorial-control markets we track the daily map published by the Institute for the Study of War (ISW) via its public ArcGIS service. We combine three layers as the basis for a market resolving — assessed Russian control, recent advances, and gains in the past 24 hours. A fourth layer, assessed infiltration, is deliberately excluded from resolution because Polymarket itself does not count it — we track it only as an early-warning proximity signal.

For each of the 50+ front-line markets we monitor, we maintain a precise geography — a point or a zone drawn around the exact city, settlement, or intersection the market resolves on — and compute the closest distance from that geography to the controlled area, assigning a threat level: CRITICAL (the zone already sits inside the controlled area), HIGH (under 500 m), MEDIUM(under 1.5 km), LOW (under 3 km), or SAFE.

To stay current without hammering ISW's servers, we ask only “has anything changed?” every few seconds during ISW's daily publish window and pull the heavy map geometry only when a layer actually moves. The result feeds the Territory tracker, Telegram alerts, and a Premium proximity API.

10What we do not claim

OrcaLayer is a business-intelligence tool over public blockchain data. It is notfinancial, investment, legal, or tax advice; we are not a bookmaker, exchange, or trading platform, and we never hold funds or execute trades. A wallet's past performance does not guarantee future results — a high win rate is a record, not a prediction.

We are also candid about data limits: some early or delisted Polymarket markets carry incomplete on-chain metadata that cannot be recovered after the fact, and reconciled figures depend on the availability of closed-position data. Where we are uncertain, we show our own computed value rather than guess. See the full Risk Disclosure and the FAQ for more.