A prediction market contract is a promise to pay $1 when something is true. Everything interesting, and every dispute worth knowing about, happens at the moment somebody has to decide whether it is.
Resolution is the least glamorous part of this category and the part most likely to cost you money in a way you did not anticipate. Two venues can list what looks like the same market, word the rules differently, and settle in opposite directions. Here is how it actually works.
The two models
Centralised resolution. Kalshi, as a CFTC-designated contract market, resolves its own contracts according to rulebooks filed with the regulator. Each contract names its settlement sources in advance — the Associated Press, ESPN, a government data release — and the exchange applies them. If you dispute the outcome, the escalation path runs through the exchange and ultimately the CFTC.
Decentralised resolution. Polymarket settles through an oracle. Outcomes are proposed, and anyone can dispute a proposal by posting a bond, which sends the question to a vote by token holders. The system is designed so that lying costs more than telling the truth, and in practice the overwhelming majority of markets resolve without anyone contesting anything.
Neither model is obviously better. Centralised resolution is faster and has an accountable party; decentralised resolution has no single entity that can decide against you for its own reasons. Both have produced contested outcomes.
What actually goes wrong
Almost never fraud. Almost always ambiguity.
The event happens, but not the way the rules describe. A contract on whether a bill becomes law by a date. The bill passes, the signature comes a day late. The rules said "by December 31" and the president signed on January 2. Everyone who was right about the substance is wrong about the contract.
The source disagrees with reality. A contract resolves on a specific data release. The release comes out, then gets revised a month later. The contract settled on the first print, because that is what the rules said.
The question was never precise. "Will there be a recession in 2026?" sounds clear until you have to name the arbiter. NBER declares recessions with a lag of many months. A contract using two consecutive quarters of negative GDP is a different bet with a different answer.
Cancellation and 50-50. Sporting contracts routinely specify what happens if a match is abandoned, a player withdraws, or a walkover occurs. Many resolve 50-50 rather than voiding, which means a position you thought was nearly settled returns half your stake.
Why this breaks arbitrage
If you trade one venue, resolution risk is an occasional annoyance. If you trade across two, it is the thing that turns a locked position into a total loss.
The arbitrage logic is that buying YES on one venue and NO on the other guarantees $1 back however the event resolves. That holds only if both venues resolve the same way. When they do not, you lose both legs at once — not break even, not take a small loss. Both.
The two venues write their rulebooks independently. They can use different sources, different cutoff times, and different treatment of edge cases, on questions whose titles read identically. This is why our arbitrage scanner tells you to check the resolution criteria on both sides before committing, and why we will not claim a spread is risk-free.
Before trading a cross-venue pair, read both rulebooks and specifically compare: the named settlement source, the exact deadline including timezone, and what happens if the event is cancelled or ambiguous. If those three match, the arbitrage is close to what it appears. If any differ, you are taking a position on the rules rather than the event.
How to read a rulebook
It takes two minutes and almost nobody does it.
Find the settlement source. Who decides? A named organisation is good. "Credible reporting" is a judgement call.
Find the deadline, with its timezone. Markets resolve on clocks, and the clock is rarely the one you are looking at.
Find the edge cases. What happens on cancellation, postponement, a tie, a withdrawal, a revision to the underlying data. This is where the surprises live.
Ask what the contract does not cover. A market on "will X be confirmed" is not a market on "will X be nominated", and the gap between those has cost people money.
The specific traps
Long-dated political contracts. Markets running years out accumulate edge cases nobody anticipated when the contract was written. Circumstances change faster than rulebooks.
"By date" versus "on date". These are different contracts. A market on whether something happens by a deadline includes everything before it; on means that day only.
Compound questions. "Will the Fed cut and inflation fall below 3%?" has four possible states and only one pays. Each additional condition roughly halves the chance while the price rarely halves to match — the same problem our parlay calculator makes visible.
Markets on other markets. Contracts referencing another platform's outcome inherit that platform's resolution risk on top of their own.
When you think a resolution is wrong
Read the rulebook again first, because most of the time the rules were followed and the expectation was mistaken. That is an uncomfortable answer and usually the correct one.
If the rules genuinely were not applied, the path differs by venue. On Kalshi, raise it with the exchange; it is a regulated entity with an obligation to apply its own filed rules, and the CFTC sits above that. On Polymarket, the dispute mechanism is built in — a proposed resolution can be challenged by posting a bond within the challenge window, which is short. Know the window before you need it.
Either way, document the contract text as it stood when you traded. Rulebooks can be clarified.
What this means in practice
Resolution risk is not a reason to avoid prediction markets. It is a reason to read before you trade, particularly on anything ambiguous, long-dated, or spanning two venues.
The habit is cheap: two minutes with the rulebook before the first trade on any new market type. The alternative is discovering the edge case after it has resolved against you, which is how most people learn this.
For how contracts price and settle more generally, start with how prediction markets work. For the venues' own approaches, our Kalshi and Polymarket reviews cover each in detail.
Frequently Asked Questions
Who decides how a prediction market resolves?
It depends on the venue. Kalshi resolves centrally using settlement sources named in rulebooks filed with the CFTC. Polymarket uses an oracle where outcomes are proposed and can be disputed by posting a bond, with token holders voting on contested cases.
Can Polymarket and Kalshi resolve the same event differently?
Yes. They write rulebooks independently and can use different sources, deadlines and edge-case handling on questions that read identically. This is the main risk in cross-venue arbitrage, because a divergent resolution loses both legs rather than breaking even.
What happens if a game is cancelled?
Whatever the rulebook says, which varies. Many sports contracts resolve 50-50 on cancellation, abandonment or walkover rather than voiding, so a position you thought was settled returns half your stake.
How long does resolution take?
Usually hours to days after the event. Contracts on data releases wait for the release; contested resolutions on an oracle-based venue take longer because the dispute window and vote have to run.
Can I dispute a resolution I think is wrong?
On Polymarket, yes — the dispute mechanism is built in, but the challenge window is short, so know it in advance. On Kalshi, raise it with the exchange, which operates under CFTC oversight. In both cases, check the rulebook first: most disputed resolutions turn out to have followed the rules as written.
What is the most common resolution mistake traders make?
Trading the headline instead of the contract. A market's title is a summary; the rulebook is the agreement. The gap between them is where money is lost.
