Pairs of Polymarket accounts trading with each other supplied at least half the volume in 928 of 1,351,870 markets reviewed by MSR. In those 928 markets, the qualifying trades accounted for $506.4 million, or 96.4% of measured volume. A large volume figure in these markets could therefore reflect repeated trading by the same pair rather than activity spread across many participants. Counts and dollar totals
The study covers April 28 to September 9, 2026, and excludes U.S. elections. MSR screened for short episodes in which the same two accounts traded with each other, finished near their starting holdings and lost little money between them. The article calls trades that met this screen qualifying trades. What qualifies and how volume is counted
How two accounts can trade $60,000 and lose $140
In the Ruiz Velasco arrest market, one account bought 30,010 Yes shares at 96 cents while another account bought 30,010 No shares at four cents. About two minutes later, the two accounts sold: the Yes shares at 26 cents and the No shares at 74 cents. Their trades added $59,780.04 to the market’s volume, yet together the two accounts were out only $139.98 after fees. Finding and definitions
The two accounts could trade so much for so little because of how paired shares are priced. A Yes share and a No share together cover either outcome and are backed by $1. In the Ruiz Velasco arrest market, the Yes and No prices added up to $1 both when the accounts bought and when they sold. As the prices moved, one account’s loss and the other account’s gain both grew large, but the two offset each other, and both accounts finished with no position. A small combined cost can therefore hide a large loss in one account and a large gain in the other. Underlying example
Large trading volume, small combined cash cost
Recorded Ruiz Velasco episode · approximate two-minute interval
96¢ + 4¢ = $1 · 26¢ + 74¢ = $1
Rare across markets, large in dollars
The 928 markets in the headline are those in which qualifying trades supplied at least half the volume. Together, these markets held $506.4 million of qualifying trades, 96.4% of their combined measured volume. Across every market in the study, including those below the halfway mark, qualifying trades came to $540.6 million, about 2.2% of the $24.4 billion reviewed. The difference between the two totals, roughly $34.2 million on the rounded figures, sits in markets outside the headline group. Counts and dollar totals
The two measures answer different questions. The market count shows how often the pattern dominated an individual market; the dollar share shows how much of all trading it supplied.
Keep the market share separate from the dollar share
April 28–September 9, 2026 · U.S. elections excluded
| Question | Calculation | Result |
|---|---|---|
| How many markets met the 50% threshold? | 928 / 1,351,870 | 0.0686% ≈ 0.07% |
| Qualifying dollars inside those markets | Published census subtotal | $506.4m |
| Qualifying dollars in the full study | $540.6m / $24.4bn | ≈2.2% of volume |
The market count also depends on where the line is drawn. Counting any market with a qualifying episode gives 8,709 markets, about 0.64% of the study. Requiring at least 90% of volume gives 531. The headline 928 uses the fixed 50% threshold.
Larger markets were more likely to cross that threshold. Of the 853,171 markets with less than $1,000 of volume, 138 qualified, about 0.016%. Of the 2,996 markets with $1 million or more, 50 qualified, about 1.67%.
Stricter and looser market thresholds
Same study frame and episode rule
| Market-level condition | Markets | Share of all markets |
|---|---|---|
| Any qualifying episode | 8,709 | 0.64% |
| At least 10% of volume | 1,558 | 0.12% |
| At least 25% | 1,198 | 0.089% |
| At least 50% · headline rule | 928 | 0.069% |
| At least 50% and $1,000 qualifying | 776 | 0.057% |
| At least 50% and $10,000 qualifying | 568 | 0.042% |
| At least 90% | 531 | 0.039% |
Flagged share was higher in larger-volume groups
Market volume during the study window
| Volume bracket · USD | All markets | Flagged | Rate |
|---|---|---|---|
| Under 1,000 | 853,171 | 138 | 0.016% |
| 1,000–10,000 | 333,267 | 209 | 0.063% |
| 10,000–100,000 | 135,440 | 278 | 0.21% |
| 100,000–1m | 26,996 | 253 | 0.94% |
| 1m or more | 2,996 | 50 | 1.67% |
A few events and a few accounts
Seven markets on Ethiopia’s next prime minister contained about $284 million of qualifying trades, more than half of the headline group’s $506.4 million. In those seven markets, qualifying trades supplied 99.6% of all volume. Other large clusters involved world leaders leaving office, Wimbledon, entry into Iran, the World Cup’s top scorer and a possible Russian nuclear test. Event groups
The accounts involved were concentrated too. Ten accounts took part in trades representing 81% of qualifying dollars, spread across 163 markets, and three of those ten took part in trades representing half. The largest single account took part in $147.9 million of qualifying trades across 24 markets. Because the listed accounts sometimes traded with one another, their shares overlap and cannot be added together. Account concentration
Ethiopia markets held the largest concentration
Qualifying trading in selected event groups · USD millions · common zero baseline
A census this large depends on getting the underlying trades right. MSR recovered 463,018,840 fills from on-chain exchange events, keeping every maker leg. Rerunning the reconstruction with different block boundaries produced the same count. A comparison with 2,617 markets from an earlier probability sample matched fill for fill and dollar for dollar, and a separately written exact-arithmetic program reproduced 30 randomly chosen qualifying markets. Reconstruction checks and coverage
For a reader looking at a Polymarket volume figure, the census gives two answers. The defined pattern dominated only a small share of the markets reviewed. Where the pattern dominated, it could supply nearly all of the volume, and a handful of events and accounts carried most of the dollars.