/leaderboard

GET https://data.vike.io/v1/polymarket/leaderboard

Polymarket wallets ranked by realized P&L, hit rate, volume or trade count, for all time or over a window, optionally inside one market category. It is the board vike.io/polymarket shows, served as an array of rows from the hourly leaderboard table, not from the raw tape. It is the same answer the polymarket_leaderboard MCP tool gives — the route runs that tool.

# the default board: lifetime realized P&L, largest first, wallets with at least 20 resolved markets, 20 rows
curl -H "X-API-Key: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/leaderboard"

# the last 30 days in crypto markets, ranked by the P&L of markets that resolved inside them, with small-trade bots left out (an average trade of at least $20)
curl -H "X-API-Key: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/leaderboard?category=Crypto&window=30d&sort=realized_pnl_window&min_avg_usd=20&limit=2"

# the biggest losers of all time
curl -H "X-API-Key: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/leaderboard?sort=realized_pnl_asc&limit=10"

# the most active by resolved markets, second page of 20: offset is rows to skip and rank continues at 21
curl -H "X-API-Key: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/leaderboard?sort=resolved&limit=20&offset=20"

# thin records too: with min_resolved=0 a wallet with two resolved markets can appear, and its hit_rate means little
curl -H "X-API-Key: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/leaderboard?min_resolved=0&sort=last_seen&limit=20"

# the board is rebuilt once an hour, so a research series is a stored one: keep each response with its timestamp
curl -s -H "X-API-Key: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/leaderboard?limit=50" -o "leaderboard-$(date -u +%Y%m%dT%H).json"

A real response for the second request, cut to its first row.

[
  {
    "rank": 1,
    "proxy_wallet": "0x32ed2e546b187ca15e2841edc82b22c713cf8ec3",
    "pseudonym": "",
    "display_name": "",
    "realized_pnl_usd": 486847.83,
    "realized_pnl_window_usd": 106260.86,
    "resolved_markets": 93972,
    "hit_rate": 0.544339,
    "hit_rate_ci": 0.003184,
    "volume_all_usd": 12839471.17,
    "volume_window_usd": 3092527.93,
    "trades_all": 599277,
    "trades_window": 112219,
    "avg_trade_usd": 21.42,
    "markets": 94080,
    "first_seen": "2026-06-11T18:16:39.000Z",
    "last_seen": "2026-10-19T15:00:15.000Z",
    "tags": ["whale", "hot"]
  }
]

Reading it

• Money is USD, ratios are fractions. realized_pnl_usd is lifetime P&L realized on resolved markets. realized_pnl_window_usd, volume_window_usd and trades_window cover window, and equal the lifetime fields when window=all. An open position contributes nothing until its market resolves, and a market that resolved invalid is excluded from P&L, hit rate and the resolved count. USD is rounded to the cent and ratios to six places.
• hit_rate is a fraction from 0 to 1, volume-weighted, and null where undefined. 0.62 is 62%. It is the share of the wallet's resolved dollar volume that sat in markets where it made money, not a count of wins. A wallet with no resolved volume has hit_rate: null, never 0; a real 0.0 means every resolved dollar sat in a losing market.
• A high hit_rate is not skill by itself. A wallet that buys near-certain outcomes just under $1 wins nearly every market for a sliver of profit, so hit_rate: 1.0 beside a small realized_pnl_usd on a large volume_all_usd is an arbitrage pattern. Sorting by hit_rate puts these first; set the P&L against the volume before reading a rate as an edge.
• hit_rate_ci is a plus-or-minus half-width, as a fraction. It is the half-width of the 95% Wilson interval, with hit_rate treated as a proportion over resolved_markets: 0.544339 with 0.003184 reads as 54.4% give or take 0.3 points. Because hit_rate is volume-weighted rather than a count, the interval is indicative, not exact. It is null where hit_rate is. A wide one means a thin record.
• The default asks for 20 resolved markets, and the top of a category can still be bots. min_resolved defaults to 20 so a rate rests on a real sample; 0 lists every wallet. Even so, the wallets with the most resolved markets in Crypto carry hundreds of thousands of them at an average trade of cents to about ten dollars: on 2026-10-19 the five at the top of category=Crypto&sort=resolved were all tagged bot. tags holds bot (an average trade under $20 over at least 100 trades), whale (lifetime volume of $100,000 or more), and hot or cold (lifetime P&L of +$1,000 or −$1,000 over at least five resolved markets). They are rules on these figures, not a classification of anyone. min_avg_usd=20 leaves out every bot.
• first_seen is not an age. It is the wallet's first trade in the recorded tape, which starts on 2026-01-01. A wallet that traded earlier, or went quiet and came back, shows a later date, so do not call one new on the strength of it. The website's new tag is not served for that reason.
• Windows are rebuilt hourly and can be revised. 24h, 7d and 30d are hour-granular, and a market's resolution time is corrected retroactively when the chain shows a different close than the one first recorded, so a windowed figure can move between two calls. The lifetime figures and window=all do not depend on a resolution time. Read a windowed rank as of the hour you fetched it.
• Paging is by offset on a board that moves. offset skips rows (0 to 1000) and rank continues across pages. Ties break on the wallet address, so one request returns one page. The board is rebuilt hourly, so a page fetched after a rebuild can repeat or skip a wallet; for a snapshot, fetch within one hour. An empty array means no wallet passes the filters.
• Wallet addresses are public on-chain data, and pseudonym and display_name are Polymarket's own profile fields, empty for many wallets. Nothing here identifies a person.
• It sorts the whole board on every call. Expect a second or so, not milliseconds, and cache the answer: it changes once an hour. For one wallet's equity curve and positions use /wallets/{address}.

Errors