How it works
Methodology
How Settlis works out a fair price, what each exchange charges to trade, what the Lab measures and refuses to measure, and how old every number is.
Last updated
The free pages are being released in stages. Where a page described here is not out yet, this is how it will work when it is.
The Settlis fair price
A contract on a prediction market pays $1 if the outcome happens and nothing if it does not. Its price, in cents, reads as a chance: a price of 62¢ says the market puts the chance at about 62%.
The Settlis fair price is Settlis's own estimate of that chance for the YES side, written the same way. It is one number for each market, in whole cents.
Where it comes from
Settlis starts from sportsbook quotes for the same outcome. A sportsbook builds a margin into its prices, so the chances implied by the two sides of a market add up to more than 100%. Settlis removes that margin, so that the two sides add up to exactly 100%. What is left is the fair price.
Every fair price carries one of two labels:
- Sharp book. A sharp sportsbook is one that takes large bets from professional bettors and moves its prices quickly. When one quotes both sides of exactly the same market, Settlis uses that pair.
- Consensus. Otherwise, Settlis combines the prices of several sportsbooks, each with its margin removed.
The sharp book comes first. Consensus is used only when there is no sharp pair.
What is published, and what is not
Settlis publishes the fair price as a whole number of cents, with its label. That is all.
Settlis does not publish sportsbook odds. The pages never show the quotes behind a fair price, which sportsbooks they came from, how many there were, how old they were, or exactly how they were combined. The fair price is a number Settlis derives. It is not any sportsbook's price.
Settlis also runs its own models. Their estimates are not published here. When a market has no usable sportsbook price for both sides, the page says there is no fair price and gives a short reason, for example that no sportsbook quotes it or that only one side is quoted. It does not fill the gap with a guess.
What it is not
A fair price is an estimate. It can be wrong, and it goes out of date as news arrives and prices move. Information only; not advice. It is not a recommendation to buy or sell anything.
One snapshot a day
The Today board on Fair prices is built once a day, as of 14:00 UTC. That is 10:00 a.m. Eastern Time in summer and 9:00 a.m. in winter. Every page prints its own as-of time, and that stamp is the one to trust.
The board is not updated during the day. By the time you read it, prices on the exchanges have moved. Never treat a number on it as a live quote.
For each market the board shows:
- the Settlis fair price and its label, or the reason there is none;
- each exchange's best bid and best ask at the as-of time. Prices only, never sizes;
- each exchange's fee, the all-in cost, and the best place to buy YES and to buy NO.
The board covers every market Settlis maps across Kalshi, Polymarket and Novig that starts within about the next 36 hours. At launch that means the NFL and soccer. It is not every market on every exchange, and some mapped markets have no fair price.
Novig prices are captured only for events that start within about 18 hours. Before that, the board says “not yet captured”.
Nothing is published about a market while its event is in play.
Fair prices published at the start
For some groups of markets, Settlis publishes its fair price only once the event starts. Until then the board shows the exchange prices and fees and says “fair published at start”.
The choice is made for a whole league and market type at a time, never for one market. A gap on the page therefore does not point at any single market.
Settlis does this to protect its own trading. Settlis also trades these markets, and its orders may be among the prices shown. Its own orders, positions and sizes are never shown and never marked. The Disclosures page explains the conflict.
Yesterday's record
The next day, at about 16:00 UTC, Yesterday's record lists every market from the previous snapshot that has since started: its fair price as of the snapshot, including the ones held back until the start; the last traded price on each exchange before the start; and the result. It also shows how well the fair prices matched the results, by league and by label, with the misses included.
What it costs to trade
Fees are charged by the exchange, not by Settlis. The tools on this site are free.
All three exchanges use the same shape of fee:
fee = rate × contracts × P × (1 − P)
P is the price in dollars, so a 50¢ contract has P = 0.50. The fee is highest at 50¢ and falls towards zero as the price nears 1¢ or 99¢. The rate depends on the exchange and on your role in the trade:
- a taker trades at once, against an order already resting in the book;
- a maker places an order that rests in the book until someone trades against it.
Kalshi
Taker. Rounded up to the next cent, for each order:
fee = 0.07 × contracts × P × (1 − P)
Maker. No fee on a series the exchange lists as free for makers. On a series whose fee type includes maker fees, rounded up to the next cent for each order:
fee = 0.0175 × contracts × P × (1 − P)
- Because the fee rounds up on every order, any order that pays a fee pays at least 1¢. One contract bought at 50¢ as a taker pays 2¢.
- Some series carry a fee multiplier that scales both fees. The examples use a multiplier of 1.
- Settlis uses the fee type and multiplier the exchange lists for each series. Where it cannot confirm them, it assumes the higher fee.
| Price | 10¢ | 25¢ | 50¢ | 75¢ | 90¢ |
|---|---|---|---|---|---|
| Taker | $0.63 | $1.32 | $1.75 | $1.32 | $0.63 |
| Maker | $0.16 | $0.33 | $0.44 | $0.33 | $0.16 |
The Maker row is a series with maker fees. On a series that is free for makers, the maker pays nothing.
Polymarket US
Taker. Rounded to the nearest cent; an exact half cent rounds to the even cent. The rate has been 0.0695 since September 17, 2026, and was 0.06 before that:
fee = 0.0695 × contracts × P × (1 − P)
Maker. No fee. The maker is paid a rebate, credited when the trade happens and rounded the same way. The board shows it as a negative fee:
rebate = 0.0125 × contracts × P × (1 − P)
| Price | 10¢ | 25¢ | 50¢ | 75¢ | 90¢ |
|---|---|---|---|---|---|
| Taker | $0.63 | $1.30 | $1.74 | $1.30 | $0.63 |
| Maker | −$0.11 | −$0.23 | −$0.31 | −$0.23 | −$0.11 |
A negative fee is a rebate paid to the maker.
Novig
Maker. No fee.
Taker. No fee on a trade made before the event starts. Novig charges takers only on trades made while the event is live, and Settlis publishes nothing in play, so the board shows a fee of zero.
| Price | 10¢ | 25¢ | 50¢ | 75¢ | 90¢ |
|---|---|---|---|---|---|
| Taker | $0.00 | $0.00 | $0.00 | $0.00 | $0.00 |
| Maker | $0.00 | $0.00 | $0.00 | $0.00 | $0.00 |
For a trade made before the event starts.
How the board uses the fees
- Fees are quoted for each contract on a reference order of 100 contracts. A smaller order can pay more for each contract on Kalshi, because of the rounding.
- The all-in cost is the price plus the fee. Buying YES at a 50¢ ask on Kalshi as a taker costs 50¢ + 1.75¢ = 51.75¢ a contract.
- The best place to buy a side is the exchange with the lowest all-in cost, among those with a captured price. It is arithmetic on price and fee and nothing else. A tie is shown as a tie.
Volume discounts, promotions and incentive programs that the exchanges may offer are not included.
Exchanges change their fees. Each exchange's own published schedule is the authority. This page states the schedules Settlis uses as of October 13, 2026.
The Lab: what it measures and what it refuses
The Lab is a nightly analysis of trades that have already happened. It covers the last 90 days of trading, only trades made before each event's start, and it is published at about 15:30 UTC. It is not a record of finished markets only: some exchanges' trades are collected as they happen, so the window can include trading on events that have not started yet.
Flow
Flow is how much trading there is in a kind of market: the contracts traded on each day of games, by league, exchange, market type and time before the start (more than 72 hours, 24 to 72, 6 to 24, 1 to 6, and the last hour). Large block trades are counted apart from ordinary ones and shown as a share.
Every flow figure carries a label that says how firm it is:
- At least, for Kalshi. Settlis's collection of Kalshi trades is incomplete, so the true figure is at least what is shown.
- At most, for Polymarket. The figure comes from summary bars, which cannot separate block trades from ordinary ones, so the true figure for ordinary trades is at most what is shown.
- Measured, for Novig. The figure is counted from the trades themselves.
Crossing cost against fair
Crossing cost is what it has cost to trade at once, taking the price on the screen. For each trade it is the Settlis fair price at that moment, minus the price paid, minus the exchange fee, in cents a contract. A negative number is a cost.
Each figure comes with a 95% range. Trades in the same event move together, so the range is worked out event by event, not trade by trade. That makes it wider, and more honest.
The Lab also lists what moved: groups whose crossing cost changed by a cent or more between two nightly versions.
When the Lab refuses to give a number
A number is shown only when at least 20 separate events stand behind it. Otherwise the cell shows the reason in place of a number:
- Too few events. Fewer than 20 events stand behind it.
- Start time unknown. The true start of the event could not be confirmed, so trades made before the start cannot be told from trades made during play.
- No fair price. There was no fair price at the time of the trades to compare them with.
- Nothing independent to compare with. The only yardstick was the market's own closing price or its result. Measuring a price against itself says nothing.
- Fair price too sparse. Too few of the trades had a fair price recorded at the same moment.
- Off the main line. The market is an alternate line away from the main one.
- Not measured. The Lab could not measure it reliably, for example because the fee could not be confirmed.
What the Lab does not do
- It measures trades that happened. It does not simulate what would have happened to an order of yours.
- It publishes nothing about trades made during play.
- It publishes no single trades, and nothing about Settlis's own orders or results.
- When Settlis corrects a definition or finds an error, the Lab says so on the page, with the date.
How old the numbers are
None of it is live. Each page is built on a fixed clock:
- Fair prices, Today
- Built once a day, as of 14:00 UTC. Not refreshed until the next day.
- Yesterday's record
- Published the next day, at about 16:00 UTC, for markets that have started.
- The Lab
- Published nightly, at about 15:30 UTC, from trades made before each start. Each exchange's data has its own age, shown at the top of the Lab.
Every page prints when its data was built. If a run is late or fails, the page says so: it shows the age of what it has, or that the data is unavailable. It never shows a zero in place of a missing number. In the Lab, each exchange's data has its own age, and any exchange whose data is stale is flagged.
Paper trading
Paper trading uses no real money and places no orders. Its results are worked out from public records of trades on the exchanges.
That has limits. Settlis cannot know where a resting order would have stood in the queue, or whether it would have traded at all, so a result is given as a range and not as a single figure. Fees are applied by the formulas above. Paper results are hypothetical.
The full method will be set out here when paper trading opens.
Limits and corrections
Coverage is partial. Some markets are not mapped, some have no fair price, and the price from one exchange can be missing from a snapshot. The pages say so when it happens.
Numbers can be wrong. If one looks wrong to you, write to support@settlis.com with the page and the market.