Primer
How to price event contracts: probabilities, fees and time
A price of 63¢ does not quite mean 63%. Here is how to turn an event-contract price into a probability, and a probability into a price.
The one-line answer
An event contract pays $1 if an outcome occurs and $0 if it does not. If you believe the outcome has probability p, and you ignore fees, interest and risk preferences, the contract is worth p dollars. A price of 63¢ therefore implies a probability of roughly 63%.
Every word after "roughly" is where the work is. The gap between the quoted price and the true probability comes from the bid-ask spread, fees, the time value of the collateral, systematic biases in who trades what, and the exact wording of the contract. A professional pricing process deals with each in turn.
From quotes to an implied probability
Markets show a bid and an ask, not a single price. The simplest estimate of implied probability is the midpoint. In thin markets the midpoint is noisy, and a size-weighted mid (or "microprice") is usually better:
The intuition: if there is far more size bid than offered, the next trade is more likely to lift the offer, so the fair price sits closer to the ask. For contracts that also trade as Yes/No pairs, check that the Yes ask and the No bid are consistent (Yes ask ≈ 100 − No bid). Gaps between them are information about where liquidity sits.
The binary option analogy
Financially, an event contract is a cash-or-nothing binary option on an outcome. For contracts whose outcome depends on a continuous quantity, such as "Will CPI year-on-year print at or above 3.0%?" or "Will the index close above 6,000?", that analogy is literally useful. If you have a distribution for the underlying quantity, the contract price is the probability mass beyond the strike.
This is why a ladder of contracts at different strikes ("CPI ≥ 2.8%", "≥ 2.9%", "≥ 3.0%"…) contains a full forecast distribution. Prices must be monotone across strikes, and the differences between adjacent strikes are the market's probability for each bucket. When a ladder violates monotonicity, something is mispriced or a rule differs between strikes.
For contracts about genuinely discrete events (elections, rulings, match results) there is no underlying continuous quantity, and the price is simply a probability estimate to be modelled directly.
Fees change the break-even probability
If a venue charges a fee on each trade, the probability at which a trade breaks even is no longer the price. Suppose you buy Yes at price P and pay a fee f per contract. You profit only if
Some regulated event exchanges have published fees that scale with P(1−P), which makes fees largest in the middle of the range and small near the tails. Where fees are rounded up per order to a whole cent, small orders pay proportionally more. The practical consequence: a market whose quotes look 2¢ too high might be fairly priced for a large trader and overpriced for a small one.
Time value of collateral
Buying a Yes contract at 63¢ ties up 63¢ until resolution. If resolution is a year away and you could earn 4% elsewhere, the fair price is lower than the probability:
For a 63% outcome a year out at 4%, that is roughly 60.6¢. That is a meaningful difference that is often mistaken for "market bias". Some venues pay interest on balances or positions, which offsets part of this effect. Long-dated markets should always be read net of carry.
Systematic biases: longshots and favourites
Decades of research on betting markets document a favourite-longshot bias: low-probability outcomes tend to be overpriced and high-probability outcomes slightly underpriced. Explanations include risk-seeking preferences among participants, the cost of capital at the tails, and the fact that being short a 3¢ longshot ties up 97¢ of collateral to earn 3¢.
Prediction markets show the same pattern in varying degrees. For pricing, the implication is that a raw 3¢ quote may correspond to a true probability nearer 1–2%, and a 97¢ quote to something closer to 98–99%. Calibrating these adjustments requires a large sample of resolved contracts. See our note on calibration.
Resolution risk: read the rules
The final adjustment is the one most often skipped. Every contract resolves according to written rules: a named source, a time window, definitions for edge cases. "Will X win the election?" might resolve on a network call, on certification, or on inauguration, and those can differ. Economic contracts may resolve on the first print and ignore later revisions, or the reverse.
A professional price is therefore a probability about the contract's resolution, not about the real-world event in the abstract. Where the rules leave room for interpretation, the price should reflect the probability of each interpretation being applied. We maintain a dedicated library of resolution criteria for exactly this reason.
A worked example
Take a contract that pays $1 if a monthly inflation print comes in at or above 3.0% year-on-year, resolving on the first official release in six weeks. The order book shows 41¢ bid for 2,000 contracts and 43¢ offered for 500.
Market-implied probability. The simple mid is 42¢. Because there is four times as much size bid as offered, the microprice is (41 × 500 + 43 × 2,000) / 2,500 = 42.6¢. The market is leaning slightly towards Yes.
Model probability. Suppose our nowcast puts the print at 2.93% with a standard deviation of 0.11 points, roughly normal. The probability of a print at or above 3.0% (allowing for rounding to one decimal place, so effectively ≥ 2.95%) is 1 − Φ((2.95 − 2.93) / 0.11) ≈ 1 − Φ(0.18) ≈ 42.9%.
Carry. Six weeks at a 4% annual rate is about 0.46%. The discounted fair price is 42.9 / 1.0046 ≈ 42.7¢.
Resolution. The contract resolves on the first print, not on revisions, so revision risk is irrelevant here. The rounding convention, however, matters a great deal: had the rules said "strictly above 3.0%", the effective threshold would move to 3.05% and the probability would fall to about 18%. One word in the rules is worth 25 points of price.
Decision. Model fair value (42.7¢) and microprice (42.6¢) agree closely. Buying at the 43¢ ask has negative expected value before fees, and so does selling at 41¢. The right action is not to trade but to quote, perhaps 42 bid / 43.5 offered in modest size, and to widen sharply in the minutes before the release, when the nowcast's uncertainty becomes the whole story.
A pricing checklist
- Start from a model probability for the outcome as defined by the rules.
- Adjust for resolution-wording risk and any known edge cases.
- Discount for the time value of collateral to resolution, net of any interest paid.
- Adjust for systematic biases at the tails, calibrated on your own resolved history.
- Compare with the market's microprice and the prices of related contracts.
- Only then decide whether, and how much, to trade, net of fees on both legs.
Frequently asked questions
Does a prediction market price equal the probability?
Approximately. A price of 63¢ implies about a 63% probability, but fees, the bid-ask spread, interest on collateral, the favourite-longshot bias and resolution wording all create small, predictable gaps.
How do you calculate implied probability from a prediction market?
Take the midpoint of the best bid and ask, or a size-weighted microprice in thin markets, and divide by the payout (usually 100¢). Then adjust for fees and time to resolution if precision matters.
Are event contracts the same as binary options?
Economically they are cash-or-nothing binary options on an outcome. Regulation, venue rules and resolution mechanics differ by jurisdiction and platform.
Related: Market making · Calibration · Glossary