Polymarket Slippage

A market order on Polymarket walks the ladder: if the top level rests a few thousand shares and your order is larger, the marginal price steps through the one-tick grid (0.5 cent steps, 1 cent deeper). The measured median trade of 42,254 shares versus top-of-book depth explains why large market orders routinely take multiple levels.

One-sidedness compounds it: with 90%+ of frames resting size on a single side, the executable ladder is often just one level deep. A buy crossing an empty ask side pulls the far bound, not liquidity — the price impact is a reprice, not a book crossing.

Reprice spikes are the worst environment: in the 4-fill study, fills slipped the wide end of the quoted spread when the reference moved; and a one-second ladder move of 99.6 cents (0.044 to 0.996) means your stop can be filled an entire dollar away from the pre-move mid.

• Compute slippage as fill price minus the executable floor at order receipt, using the 250ms frame stamped at receipt, not the mid — the mid is often midless-execution, and the floor is what you actually pay.
• Requote the floor file before and after your fill: the archive has the frames on both sides of your timestamp, so impact is measured as the floor move, not the tick price.
• Benchmark per series: Up/Down ladders are exponent-tight and shallow; sports books are 35–36 levels deep both sides. Slip per dollar is a per-series number that the same query shape replicates.
• Track final-minute slippage separately: ask presence in 3 of 18 sampled frames tells you the last-minute ask is on vacation exactly when exit orders arrive.

Polymarket cannot offer exchange-depth liquidity, because the book is a prediction ladder, not an order book over a continuous asset. There are no hidden edges to legalize: an empty ask side is an empty ask side, and slippage models that assume a deep book will be penalized by the archive daily.

The practical mitigation is structural: post maker quotes, size to the resting side, exit with staged orders across 0.5c levels, and use the 250ms archive to compute realized slippage per strategy instead of assuming a fixed basis-point cost.

None of this is an argument against trading — it is the argument for instrument-aware execution: the same ladder that slips a last-minute exit is the same ladder that gave a 99.6-cent bound in a second, and knowing which you are holding is the whole game.

• Pick a liquid Up/Down market from the archive and replay one minute of frames around its busiest reprice.
• Simulate a 42,254-share buy at each second against the resting ladder and record the marginal fill price.
• Compare the fill series to the executable floor; the spread between them is your slippage per second.
• Run the same test on a 35-level sports book and report the difference — the two numbers are the honest baseline for execution cost per series.

Slippage on Polymarket is the price of a thin, often one-sided book priced in single ticks — and the cost is measurable, not mysterious: fill against the executable floor, measure the floor move, slice it per series.

The numbers that frame every decision are short: median trade 42,254 against shallow tops, a 99.6-cent ladder in one second, ask presence in 3 of 18 final frames. Execution design that ignores those three facts overpays by tick after tick.

Do the 60-second replay before your first live order: a simulated 42,254-share walk up the ladder, measured per second against the floor, is the cheapest education the data can sell.

• Implied probability
• Liquidity
• Order flow
• Settlement data