We think most people new to prediction markets should start by posting resting limit orders, which makes them the maker, instead of paying the ask as a taker. Three sets of evidence point the same way: the fee schedules, a study of who made and lost money on Kalshi, and our own data on how close prices land to outcomes. This is a Perspective piece. The facts below are sourced; the conclusion is the Handbook’s opinion. It explains how order types affect costs and is not a recommendation to trade.
What does a taker pay that a maker does not?
A taker pays a fee and crosses the spread. A maker usually pays neither.
- Kalshi. The taker fee is round up(0.07 x C x P x (1 - P)), which is 1.75¢ per contract at 50¢. The maker formula uses 0.0175 with a multiplier that “is 0 unless otherwise indicated” (fee schedule effective July 7, 2026). Kalshi’s fee page lists 160 series with non-standard fees, and some charge makers, including the NFL game series. Canceling a resting order is free (Kalshi help).
- Polymarket. “Makers are never charged fees. Only takers pay fees” (Polymarket fees). Taker rates run from 0 for geopolitics to 0.07 for crypto, and those fees fund daily rebates paid to makers whose orders get filled.
- Polymarket US. Accounts under $10,000 of 30-day taker volume pay Θ = 0.095 after their first week, or 2.375¢ per contract at 50¢ before rounding. Standard maker fills pay no fee (Polymarket US fees).
The spread adds to the fee. A University College Dublin paper on Kalshi walks through a real order book: an April 2025 CPI contract where the best YES offer was 32¢ and a trader could instead post a bid at 30¢. Under Kalshi’s taker formula, buying at 32¢ adds about 1.52¢ in fees, so the taker needs the event to happen about 33.5% of the time to break even. A filled bid at 30¢ breaks even at 30%. That is 3.5 points of probability on one contract. Our odds guide explains spreads and break-even in more detail, and the break-even calculator runs the numbers.
What does the research say about makers and takers?
Bürgi, Deng and Whelan (January 2026) used Kalshi’s own trade data, which records which side was the maker, on more than 300,000 prices from 2021 through April 2025. They stopped in April 2025 because Kalshi began charging some makers after that. Their findings:
- The average return was minus 9.64% for makers and minus 31.46% for takers, after takers’ fees.
- Makers who bought contracts at 50¢ or more earned 2.6% on average.
- Cheap contracts, the worst performers, were mostly bought by takers. Makers bought 43.5% of contracts priced 1¢ to 10¢ and 56.5% of those priced 90¢ to 99¢.
The authors’ explanation is that makers get better prices but take on the chance of not being matched, and “must also be willing to cancel those orders if new evidence emerges.”
What does our own accuracy data add?
Our accuracy study, computed November 11, 2026, covers 1,451 resolved Polymarket markets with at least $100,000 in volume that closed between September 12 and November 10, 2026. It compares the price one day and one week before close with how often that outcome won.
One day before close, 490 markets were priced between 40% and 60%. They averaged 49.4%, and 49.8% resolved YES, a gap of 0.4 points. Across all ten price bands, weighted by the number of markets, the average gap was 2.0 points. For comparison, a taker fee at 50¢ adds 1.75 points on Kalshi, 1.25 points on a Polymarket sports market and 2.375 points on Polymarket US at Tier 1.
Our read: in the price range that held the most markets (490 of 1,187), the market’s average error was smaller than the fee a taker pays. A taker needs to be right by more than the fee plus half the spread in each specific market. A maker filled at their own price keeps both. Our data has limits. It records hourly prices without bids and asks, so it cannot say which side was the maker. Some bands showed larger gaps: 42 markets priced 10% to 20% averaged 15.8% and resolved YES 23.8% of the time. Small bands move a lot between runs, and markets within one event are not independent.
What is the strongest case against starting as a maker?
The strongest objection is adverse selection. A resting order is a free option for everyone else. It tends to fill when news has just moved the true probability past your price, because the trader hitting it knows something you have not yet seen. The exchanges treat this as real:
- Polymarket cancels resting limit orders on sports markets at the official start time, and warns that if a game starts early, orders “may not be cleared in time” (Polymarket markets docs).
- Polymarket US cancels all open orders before weekly maintenance, because leaving them “would expose traders to stale fills” (Polymarket US FAQs).
Two more objections follow. First, fill risk: Kalshi notes “there is no guarantee a limit order will be filled” (Kalshi limit orders), and the trades a maker misses may be the best ones. Second, selection: the Kalshi paper’s model has traders with stronger beliefs choosing to take, so part of the roughly 22-point gap reflects who makes and who takes. A new trader who switches order types does not inherit the makers’ average.
We accept the selection point. The paper’s averages describe groups and are no forecast for any one person, and makers still lost 9.64% on average. Posting lowers costs, and any edge still has to come from the trader’s own view.
Two parts of the case survive the objections, though. The fee saving and the spread are mechanical: every filled maker contract saves them, whatever the trader’s skill. Adverse selection is real but concentrated around news, scheduled releases and live games, which a new trader can choose not to leave orders resting through, and canceling on Kalshi costs nothing. An unfilled order loses no money. For someone without a fast information edge, that trade-off favors posting.
What would change our mind
- Maker fees spreading. If Kalshi applied its 0.0175 maker rate to most markets, or Polymarket began charging makers, the fee gap would shrink to three-quarters of the taker fee or less.
- Post-2025 evidence. A study of Kalshi or Polymarket trades after April 2025 showing maker returns no better than taker returns would undercut the research leg.
- Fill quality data. Evidence that small resting orders from retail traders mostly fill just before adverse moves, by enough to outweigh the fee and spread saved, would answer our argument directly.
- Our own data. If future runs of our accuracy study showed mid-range price errors consistently larger than taker fees, a taker with a real view would have more room than we credit here.
For the full fee formulas and how the venues differ, see Kalshi vs Polymarket.