Partner research

Polymarket: A User-Base Portrait

Pantera Research Lab × Surf — who trades on the international exchange, how they arrive, how they behave, and how consistently they perform.

I

The two Polymarkets

Polymarket runs two venues with different plumbing. The international exchange settles on Polygon and emits the fills every on-chain figure here is built from. Polymarket US settles off-chain through FCMs and emits none of them, so the two are measured separately and never summed. This part sizes both venues, then takes the international book apart: what trades, who carries the flow, and how participation is distributed.

Monthly volume — international vs US

One-sided monthly volume by venue

Monthly traded notional on each venue, on the one-sided deduplicated convention. US figures come from the verified FCM archive, which is missing a minority of days in the window; affected months sum the days that exist and are disclosed in the handoff rather than padded with zeros.

Daily venue metrics — Polymarket US

Reported fills, taker volume and open interest per day on Polymarket US

Daily series from every official FCM archive file in the window. Dates with cross-report reconciliation warnings remain visible at file grain and are not promoted into mapped market analytics. The archive does not publish participant identity, so wallet counts are omitted rather than shown as zero.

Fee revenue and take rate by category

Reconstructed fees and effective take rate per category

Bars are reconstructed fee revenue per market category; the line is the effective take rate that revenue implies against the same category’s volume.

Volume by market category

One-sided volume per category, with each category’s largest market

Traded notional by category on the international venue. Each row names the largest market in that category and shows its volume and share of category volume.

Active wallets by side of the book

Any-side union, ever taker, and ever maker

Any side is the unique-wallet union. A wallet active as both maker and taker appears once in that union and in both side-specific bars, so the latter must not be summed.

Volume concentration by wallet rank

Share of volume by wallet rank tier — takers (demand) vs makers (supply)

Two 100%-stacked volume bars show how flow is distributed across wallet-rank tiers on the demand and supply sides of the book.

Engagement funnel

Share of wallets vs share of taker volume per activity tier

Wallets bucketed by how much they traded, showing the gap between how many wallets sit in a tier and how much of the volume that tier carries.

Venue split — binary vs multi-outcome

Standalone binaries (CTF) vs mutually-exclusive multi-outcome (Neg Risk)

One bar, split by exchange contract: how the venue’s traded notional divides between standalone binaries and mutually-exclusive multi-outcome markets, with each contract’s fill count alongside.

II

User growth on the international venue

Who arrives, what brings them in, and how much of each month’s cohort is still there the months after.

User growth accounting

Monthly active wallets by cohort — new, resurrected, retained; churn below the axis

Each month’s active wallets split by where they came from, with the wallets lost since the previous month drawn below the axis. New versus resurrected is judged against full lifetime history; the first displayed month is published only when the immediately preceding month is available for retained and churned counts.

New users by entry category

New wallets by the category of their first market, per cohort month

Each month’s newly arrived wallets, stacked by the category of the first market they ever filled. The mix of what brings people in is the point; the height is the intake.

Top onboarding markets

Markets ranked by count of new wallets whose first-ever fill was there

The individual markets that served as the front door: each row counts wallets whose first fill on the venue happened in that market.

Onboarding dispersion by month

How much of each month’s new-wallet intake the top-15 (and #1) onboarding markets capture

Concentration of the front door: the share of a month’s intake captured by its fifteen biggest onboarding markets, and by the single biggest alone.

Monthly retention — platform vs category

Month-over-month retention — any-category vs same-category

Of the wallets active in a month, the share still active the following month: once for the platform as a whole, and once per category counting only wallets that returned to that same category.

Retention by entry category

New-wallet cohorts still active 1 and 3 months later, by entry category

Of the wallets that entered through each category, the share still active one and three months on. Cohorts are pooled across the months for which each horizon is observable.

Cross-category user overlap

Shared wallets between category pairs, as a share of the smaller category

Each cell is the wallets active in both categories divided by the smaller of the two, so it reads as a cross-sell rate rather than a size comparison. The diagonal carries each category’s own active-wallet count.

III

The three tribes

Wallets are labelled by behaviour, not identity — bot-like first, then sophisticated, with retail as the remainder. The labels describe how an address trades. They are not claims about who owns it.

Specialization by category breadth

Wallets by number of categories traded (of 7) · share of wallets vs volume

Wallets grouped by how many of the seven categories they touched, with each group’s share of the population set against its share of taker volume.

Fills per wallet-market relationship

How many times a wallet trades the same market, by category

Each row splits a category’s wallet-market relationships by how many fills the wallet made in that market — one-and-done on the left, repeat engagement to the right.

How wallets are classified

Behavioral signals over each wallet’s H1 fills (both sides, deduplicated) — evaluated in order: bot first, then pro; retail is the remainder

Probable Bot

  • more than 100 fills per active day, or
  • more than 500 fills with robotically uniform sizing (CV < 0.05), or
  • more than 1,000 distinct markets

Informed/Pro

  • avg fill ≥ $500 with ≥ 5 fills across ≤ 50 markets, or
  • avg fill ≥ $200 with ≥ 10 fills across ≤ 100 markets, or
  • > $50K volume concentrated in ≤ 30 markets

Retail

  • everyone else — the default cohort

The rules are deliberately coarse and behavioral. They describe how an address trades — not who owns it.

Trader cohorts — wallets vs volume

Share of wallets vs share of taker volume per behavioural cohort

The population split set against the volume split. The distance between the two bars is the whole point of the classification.

Cohort volume share by category

Taker-volume share of each cohort within each category

Who carries each category’s flow: the share of taker volume attributable to each behavioural cohort, per category.

Cohort wallet share by category

Wallet share of each cohort within each category

The population mix behind the flow: the share of active wallets in each behavioural cohort, per category.

Trade size distribution by cohort

Taker-fill size distribution per behavioral cohort

How each cohort sizes its fills, from sub-dollar dust to five-figure clips. Shares are of the cohort’s own fill count.

Cohort volume share around fee rollout

Daily taker-volume share by cohort · dotted line = fee switch-on

The volume mix through each category’s fee introduction. The vertical line marks the day fees switched on for the selected category; World / Geopolitics had no fee event in the window.

Cohort wallet share around fee rollout

Daily share of active wallets by cohort · dotted line = fee switch-on

The population mix through the same fee events — whether the wallets showing up each day changed compositionally when trading stopped being free.

IV

Who makes money

When each behavioural cohort enters a market’s life, how often it wins, how realized profit and loss is distributed, and whether profitable months persist.

Entry timing within markets

Distribution of each cohort’s volume-weighted mean entry position · boxes p25–p75, whiskers p10–p90 · markets ≥ $100K & ≥ 50 takers

Each market contributes one volume-weighted mean position per cohort over its complete observed trading life, where 0% is the first eligible fill and 100% the last. Boxes summarize that market-level distribution, and the printed n is the number of markets behind each box. Markets are included only once resolved. The population rolls over the six most recent complete months, and segment labels are frozen to wallet behavior in that same window.

Top market per category — daily cohort volume

Each category’s biggest market — daily taker volume by cohort

The single largest market in the selected category, day by day, split by who was trading it. Days with no observed fills for a cohort are gaps, not zeros.

Realized PnL tiers — gross of fees

Wallets by realized H1 PnL tier (markets resolved in H1) · net $ per tier

Each tier’s aggregate dollar outcome around a zero line, with the tier’s wallet count and median win rate alongside. Winning tiers point right, losing tiers left.

Profitability and win rate by cohort

Wallet profitability, median wallet win rate, and pooled position win rate

The profitable share and median win rate are wallet-level. Pooled win rate is position-level, dividing all winning resolved positions by the displayed position count. Wallets without trades in the window are excluded from this three-cohort view.

Net realized PnL by cohort

Who pays whom: each cohort’s aggregate H1 PnL on markets resolved in H1

Aggregate realized profit and loss, gross of fees, on markets resolved inside the half. The three displayed trading cohorts exclude wallets with no H1 trades; that excluded group carries the balancing residual that brings the complete resolved-market ledger to zero.

Profit persistence

Wallets by number of net-positive months

How many wallets were net positive in how many distinct months, on a cash-flow basis. The amber overlay is the subset for which one month generated more than 90% of positive-month profit. The value axis is logarithmic.