Polymarket Order Book Depth: 10,000 Contracts Fit 64% of Books

The quoted price holds for 100 contracts and often not for 10,000. Every one of 2,701 Polymarket order books showed enough asks to fill 100 contracts, 78% showed enough for 1,000, and 64% enough for 10,000.

Where the order did fit, it usually cost little. The median 10,000-contract buy paid 0.25¢ a contract above the best ask, before fees, and the median 1,000-contract buy paid nothing above it; the cost sits in the thin books, where one 1,000-contract order in ten paid 9.38¢ or more.

Vultax took the first Polymarket order-book snapshot of each hour on 17 and 18 September 2026, 7,354 books in all, and walked up the displayed asks at three order sizes in every book with a best ask between 10¢ and 90¢. The markets came from a watchlist built for event research; a second read a month later compares the busiest books with markets drawn at random.

A snapshot holds up to 20 price levels a side. An order fits when those levels hold enough contracts; its cost is the average price paid walking up them, less the best ask. Each book is one side of a market's token pair.

On Polymarket the 90th-percentile cost falls from 9.38¢ at 1,000 contracts to 1.65¢ at 10,000. Bigger orders did not get cheaper: the thin books that made 1,000 contracts expensive cannot fill 10,000 at all, so they drop out of the larger order's average.

Hold the books fixed and the curve runs the right way. In the 1,717 Polymarket books that could fill every size, the median cost was 0.00¢, 0.00¢ and 0.25¢ at 100, 1,000 and 10,000 contracts. In every book, a larger order paid the same average price or a higher one.

A median of zero means half the orders filled entirely at the best ask. The other half is where the cost sits: one order in ten paid 9.38¢ or more above the best ask at 1,000 contracts.

A month later we read Polymarket again, books only, with no orders placed. At 23:09 UTC on 19 October 2026 we took one snapshot of two sets of open markets: the 400 with the most contracts traded in the previous 24 hours, and 1,500 drawn at random from all 253,337 open markets taking orders. The same rule as September applied: a positive bid below the best ask, and a best ask between 10¢ and 90¢. Of the busiest books, 186 passed, and 343 of the random ones. Most of the random draw fell out before any order was walked: 502 books had no bid at all, 413 had a best ask above 90¢ and 140 were empty.

In the busiest books, size was cheap. 10,000 contracts fit 90.9% of them within 20 price levels, at a median 0.68¢ above the best ask, and the median spread between bid and ask was 1¢. A market at random is another place. There, 10,000 contracts fit 21.9% of books within 20 levels, at a median 19.19¢ above the best ask. 1,000 contracts fit 70.8%, at 7.22¢. Even 100 contracts paid a median 0.87¢ above the best ask, and one order in ten 33.92¢ or more, because in 216 of the 343 books the best ask held fewer than 100 contracts. The median spread was 7¢.

The September watchlist, built for event research, sits between the two: 63.6% of its books could fill 10,000 contracts. A depth figure is only as good as the list of markets it was read on, and the list belongs next to the figure.

Traders think in money, not contracts, and a dollar order buys more contracts the lower the price. Walked in dollars on the whole book, $1,000 fit 99.5% of the busiest books at a median 0.16¢ above the best ask, and $10,000 fit 95.7% at 1.52¢, with one order in ten paying 26.24¢ or more. In the random draw $1,000 fit 69.1% of books at a median 18.51¢, and $10,000 fit 22.2% at 31.89¢.

The whole book holds more than the 20 levels September captured. 155 of the 186 busiest books showed more than 20 ask levels, and counting all of them lifted the share that could fill 10,000 contracts from 90.9% to 97.8%. The extra fills are the expensive ones, deep in thin books, so the 90th-percentile cost rose with them, from 5.23¢ to 12.80¢. It is the selection described under 'Why 10,000 contracts can look cheaper than 1,000', seen from the other side.

A single book shows what the averages are made of. On Giants vs. Texans, 25 October, the market on the Texans covering a 9.5-point spread, the best bid was 50¢ and the best ask 51¢ at 23:09 UTC on 19 October. Only 72.11 contracts sat at the best ask, $36.78 worth. A buyer of 10,000 contracts would have taken them, then 3,512.33 at 52¢ for $1,826.41, 3,246.86 at 53¢ for $1,720.84 and the last 3,168.70 at 54¢ for $1,711.10.

That is $5,295.12 for 10,000 contracts, an average of 52.95¢, or 1.95¢ a contract above the quote. A screenshot of the 51¢ ask would have been right for 72.11 of them.

A screenshot of a price says nothing about size. Three checks follow from the numbers.

• Match the size, not the price - 100 contracts filled at or near the quote almost everywhere; at 10,000 more than a third of the September books could not fill the order from what they displayed, and in markets drawn at random most could not.
• Add the delay - these are books frozen at one moment. Following a Polymarket trade 30 seconds late has its own measured cost.
• Add the fee - on a fee-charging Polymarket market the taker fee comes on top of the fill; the fee calculator gives it for a size, price and category.

The source is Vultax's order-book archive for 17 and 18 September 2026, from which the first recorded Polymarket cycle of each UTC hour was kept: 48 cycles covering 83 Polymarket markets, 7,354 books in all. Books enter the size test with a positive, uncrossed bid and a best ask between 10¢ and 90¢, which keeps near-settled markets from dominating the costs.

For an order of a given size, asks are sorted by price and consumed level by level; the cost is the sum of price times quantity divided by the order size. When the captured levels hold fewer contracts than the order, the order is counted as not fitting and gets no cost. Depth can exist beyond the 20 captured levels, and a displayed order can be cancelled before anyone reaches it.

The watchlist was built for event research, and hourly snapshots of one market are related to each other. Fees, order minimums and the time between seeing a price and sending an order are outside the measurement. Challenge a figure at [email protected]; material changes are logged in the changelog.

The second read of 19 October 2026 is one snapshot of Polymarket only, from the public order-book endpoint, which returns the whole displayed book. The busiest set was taken from Polymarket's market feed ordered by 24-hour volume at the time of the read; the random set was drawn from every open market accepting orders, with a fixed seed, and read on each market's first outcome as in September.

Input and derived data files are not offered for this study. The earlier files were built on a sample that included another venue and have been withdrawn rather than reissued. Methods and limitations remain described here; contact [email protected] for questions. This is not a claim that Vultax owns the underlying venue data.

Copy-trading services that want their own alerts replayed this way can write to [email protected]: send a timestamped alert log and the execution rules, and Vultax will publish the fill table at 100, 1,000 and 10,000 contracts with the fee schedule applied, misses included.