Best Polymarket Traders: 36 Wallets Took 44% of Gains; Winners Repeat 41–51% of the Time

A small number of Polymarket wallets take most of the money, and the ones that trade often keep taking it. In Vultax's record of 9,902 wallets from 18 June to 14 July 2026, the top 1% of winning wallets, 36 addresses, took 43.9% of all realised gains, while 54.3% of wallets closed nothing at all.

Winning does not usually repeat: across all wallets active in consecutive months from March 2025 to March 2026 (431 million fills, 13 month pairs), a profitable month was followed by another 41–51% of the time. The most active are the exception. Among the 184 wallets with ten or more trades in both June and July 2026, 124 of the 137 June winners were profitable again in July. Activity is the first filter: in July, one-trade wallets were in profit 12.2% of the time, against 63.7% for wallets with six to twenty trades.

No leaderboard shows any of that. Polymarket's own board ranks by profit and volume with no definition of profit on the page, and none of the five boards compared here states whether its P&L is net of taker fees. The practical order: filter on activity, then read the rank, then find out how the rank was computed.

The Polymarket trader leaderboard in the Vultax terminal ranks wallets two ways. Vi IQ, the default, orders the most profitable wallets of the last 30 days by profit, profit per dollar traded and the last 7 days, and lists 50 that traded at least $1,000,000 in that time and are up all-time. All time ranks 80,000 wallets by the lifetime profit Polymarket reports. The top 50 of each order are free to read, and so is any wallet's public record: paste its address and it opens with Polymarket's rank, lifetime P&L and volume, plus the win rate for a wallet among the 80,000. Open positions, every trade, P&L by category and every rank after the first 50 are on Essential. The board has four views, and each answers a different question.

The figures in this study come from two fixed research samples: 9,902 wallets for where the gains went, and 7,178 for activity, both measured in June and July 2026. They show how to read a leaderboard. They are not the population the live board ranks, so a percentage here should not be applied to a rank there. A wallet that is missing from a view has no record in it, which is different from a record of zero.

Download table figures:CSVJSON with sources

Vultax's prediction-market pipeline keeps, for every wallet it tracks, the realised profit and loss booked when the wallet sells, computed from its own cost basis, with trade counts and hit rates. Two readings are used here: 27,376 wallet-market records with closed trades between 18 June and 14 July 2026, covering 9,902 wallets and 2,147 markets, for where the money went; and the calendar months of June and July 2026, covering 9,949 wallets and 13,582 wallet-months, for whether one month's result predicts the next. A third, wider reading covers every wallet with a realised result in two consecutive months from March 2025 to March 2026 — 431 million fills across 13 month pairs — and is used only for the population-wide repeat rate.

Realised means closed by trading. A wallet that bought Yes at 40 cents and is still holding shows nothing, whether the market has since moved to 90 or to 5. This measures who took profits and losses off the table, not who was right.

In July, 7,178 wallets recorded at least one trade and 36.2% of them finished the month with a realised gain. That headline hides a steep gradient: 12.2% of one-trade wallets were profitable, against 63.7% of wallets with six to twenty trades and 65.9% of those with more than 100. June showed the same shape at a higher level, with 50.1% of 6,404 wallets profitable.

Part of this is measurement rather than skill: a one-trade wallet shows realised P&L only if it closed that position inside the month, and most are still holding. That is why leaderboards built on realised P&L favour active wallets, and why sample size has to sit next to rank before rank means anything.

Download chart figures:CSVJSON with sourcesSVG chart

• 1 trade — 2,199 wallets, 12.2% profitable, mean hit rate 0.12
• 2–5 trades — 3,665 wallets, 40.8% profitable, hit rate 0.21
• 6–20 trades — 1,153 wallets, 63.7% profitable, hit rate 0.27
• 21–100 trades — 117 wallets, 63.2% profitable, hit rate 0.26
• 100+ trades — 44 wallets, 65.9% profitable, hit rate 0.26

Of 9,902 wallets, 3,501 (35.4%) closed at least one position at a gain, 1,026 (10.4%) closed at a net loss, and 5,375 (54.3%) closed nothing at all. Among the winners, gains were steep. The top 1%, thirty-six wallets, took 43.9% of all realised gains. The top 10%, 351 wallets, took 79.5%. The hundred largest winners took 58.7%. One wallet alone booked $106,663, 13.6% of every realised dollar in the sample; the next four largest took $41,000, $14,310, $13,947 and $13,205.

The median wallet's realised P&L was zero; the 90th percentile was $75 and the 99th $1,212. Markets were concentrated too: the top 1% of markets, twenty-one of them, accounted for 43.8% of absolute realised P&L, and the single largest produced $107,156 of net realised P&L across 26 wallets, the great majority of it that one wallet's gain. Following a whale usually means following one wallet into one market.

Download chart figures:CSVJSON with sourcesSVG chart

Realised gains in the window totalled $783,051. Realised losses totalled $15,371. The largest loss any wallet booked was $13,809, and the next largest was $264. Fifty-one dollars of gains were realised for every dollar of loss.

That ratio is not a statement about how often traders are right. It is the disposition effect, which Odean documented in 1998: investors realise gains far more readily than losses. A binary market amplifies it, because a losing position can simply be held to resolution and expire at zero without a sale ever being booked. The losses exist; they are taken by not trading.

Counted by markets over the same four weeks, the gradient is the same: wallets that traded a single market closed a gain 19.9% of the time (4,396 wallets), two or three markets 39.9% (3,444 wallets), four to ten 60.7% (1,875 wallets), and eleven or more 58.8% (187 wallets).

Download chart figures:CSVJSON with sourcesSVG chart

184 wallets traded ten or more times in both June and July 2026. Of those, 67.4% were profitable in both months and 23.9% unprofitable in both; 7.1% went from a profitable June to an unprofitable July, and 1.6% the other way. Read from June's side, 137 of the 184 were profitable in June (67.4% + 7.1% = 74.5%), and 124 of those 137, or 90.5%, were profitable again in July. The Spearman rank correlation between their June and July realised P&L was 0.83: where a wallet sat in the June ranking was a strong guide to where it sat in July.

That is the active cohort, and it is not the population. Across every wallet with a realised result in two consecutive months from March 2025 to March 2026 — 431 million fills over 13 month pairs — winners repeated 41–51% of the time. The ten-trade filter is doing the work: persistence is a property of the active few, not of winners in general.

At the very top the effect was stronger still. Of the 24 wallets in June's top decile by realised P&L among active wallets, 79.2% were in the top decile again in July and 87.5% were still profitable. Only 12.5% fell into the bottom half.

Two months is a short window and these are small groups. Carhart's classic study found that most mutual-fund performance persistence disappears within a year and is largely explained by momentum and fees; Mellers and colleagues found the top 2% of forecasters in tournaments kept their edge across years. Over eight weeks, this sample looks more like the forecasters.

The active group is chosen with July in hand: a wallet had to trade ten times in July to be counted, so wallets that went quiet after a good June are not among the 137. A test run forwards would fix the group at the end of June and count the ones that stopped.

The dollars are modest. Among wallets with ten or more trades in a month, 275 in June and 444 in July, 66.2% and 65.8% were profitable. The median made $16 in June and $15 in July and the 99th percentile $6,825 and $6,356; total realised gains across these wallets were $150,000 and $250,000. This is a retail-scale cohort with a thin positive tail, not a room full of whales.

Hit rate separates the groups sharply. Among wallets with twenty or more trades in a month, those whose hit rate on closed positions was below 20% were profitable 27% of the time (n=152); between 20% and 40%, 96% (n=51); above 40%, 100% (n=97). Hit rate is itself shaped by which positions got closed, so this describes the sample, not a rule that a 20% hit rate guarantees a gain.

Five public boards, scored on what they rank, whether they say how, and how often they move, read from each vendor's own pages on 5 September 2026.

Download table figures:CSVJSON with sources

Three things decide whether a top trader is worth following, and a rank contains none of them: activity, concentration and basis.

Activity and concentration are measured above. A quiet wallet at the top of a monthly board is usually one resolved position, not a strategy, and a $100,000 month from one market reads the same on a board as a $100,000 month from forty.

Basis. "Profit" on a board can be realised, marked to the current price, or both; it can be gross or net of Polymarket's taker fee, which is C × feeRate × p × (1 − p) with feeRate 0.04 to 0.07 on most categories (fee docs). None of the five boards says which. Two traders with the same displayed number can be in different businesses.

A checklist that turns a rank into a decision. The same checks apply in Vultax's trader analytics, which keeps the sample and the time window beside each wallet's observed performance.

• Trade count first - discard wallets with fewer than ten trades in the window; the persistence evidence lives above that line.
• Window second - a 30-day rank is a snapshot; check the same wallet on the all-time view and on last month's dated page where one exists (Polycopy keeps them).
• Category third - specialists persist better than generalists; Polycopy's Copy Score and the Vultax category profiles both exist for this reason.
• Basis last - if the board does not say realised or marked, gross or net, assume the most flattering combination and discount accordingly.
• Never copy a rank - copying pays your lag, the taker fee and the fill the leader just moved; the copy-trading study measures that cost, and Polycopy's free paper-trading bots let you test a wallet for a month first.

The strip at the top of this page is rewritten every ten minutes from Polymarket's public leaderboard by day, week and month, its data API's trades of $10,000 or more, its global open interest, and Vultax's record of every fill it observes, including the always-on wallet cohort. The Vi IQ beneath it scores the six domains the Vultax terminal scores for a pair, computed for this page's subject from published inputs, with a domain that cannot be computed shown as unavailable rather than filled in. The figures in the body of the article are the measurement it was written on and are dated; the strip is what the same measurement says now.

The four figures are the top wallet's profit over the last seven and thirty days, the ten best wallets' combined seven-day profit, and the share of the top twenty-five's profit the top ten took. Polymarket's weekly board resets on Monday at 00:00 UTC, so the seven-day figure is small early in the week and the thirty-day one is the steadier reference; names are the public display names the venue itself shows.

Each chart in this study has a download row beneath it: the figures as CSV, the same figures as JSON with their sources, and the chart itself as an SVG image that carries its title, window and source line. The activity chart covers 7,178 wallets in July 2026 (SVG, CSV). The concentration chart covers 9,902 wallets between 18 June and 14 July 2026 (SVG, CSV).

Credit Vultax Research, link this study and state the window and the number of wallets with any figure you quote: "43.9% of realised gains went to the top 1% of winning wallets (Vultax Research, 9,902 Polymarket wallets, 18 June to 14 July 2026)". Neither chart is a current leaderboard. Keep the JSON beside a reused CSV, because it holds the sources, the method and the version of the table. Reuse follows the site's terms.

The population behind every leaderboard changed in 2026. Polymarket opened a U.S. exchange in December 2025 and launched perpetual futures with up to 20x leverage on 3 September; Kalshi's contract volume reached about 90% of the pair's combined total in late August on sports and combos, and August was the first monthly decline in a year after the World Cup summer. A leaderboard that counts profit in dollars ranks a different crowd each season.

Two findings from outside this site should sit next to any ranking. Columbia researchers flagged about a quarter of Polymarket's volume over three years as wash trading, roughly one wallet in seven, many of them with no profit at all, which is what farming an airdrop or a ranking looks like. And the CFTC's February 2026 advisory cites a candidate who traded his own race and a video editor who traded on non-public channel information: some of the 'smart money' a tracker surfaces is money that knew.

Board facts were read from each vendor's own pages on 5 September 2026. The Vultax figures are computed from public on-chain fills over two short windows and measure realised results only. Wallets present in both months are the ones that kept trading, which is a form of survivorship, and wallets enter the pipeline through leaderboard and activity discovery, so the sample leans toward visible, active traders. The wider March 2025 to March 2026 reading applies no activity filter, so it counts wallets with a single closed trade in each month; that is why its repeat rate sits well below the active cohort's.

Realised P&L depends on the pipeline's cost-basis accounting and excludes open positions (the 18 June to 14 July reading also excludes resolution payouts), so it will not match Polymarket's own profit figures. Wallet identifiers are public addresses; nothing here identifies a person, and nothing here is financial advice. Challenge a cell at [email protected]; material changes are logged in the changelog.

• Profitable — realised P&L above zero in the month; open positions excluded
• Dataset — the activity and profitability table above is available as a CSV on this page