A sports prediction market lets participants trade contracts tied to a defined outcome. The useful starting point is the payout, the executable price and the settlement rules. A displayed price is not a verified true probability, and trading on an exchange does not eliminate costs.
A hypothetical $1 contract
Suppose a YES contract pays $1 if Dallas wins a specified championship and $0 otherwise. Buying 100 contracts at 15c costs $15 before fees. A winning settlement returns $100; a losing settlement returns $0.
The gross break-even win rate is 15%. That translates the payout into a probability threshold; it does not establish Dallas’s actual chance. If the ask is 16c while the last trade was 15c, a purchase at the ask has a 16% gross break-even rate instead.
Read the contract’s rules for the event, season, settlement source, cancellation and exceptional outcomes. A superficially similar sportsbook bet may settle differently.
Comparing with sportsbook odds
A hypothetical +550 sportsbook line implies 100 / (550 + 100) = 15.38% as a gross break-even win rate. You cannot derive the sportsbook’s “true probability” from that one number. A no-vig estimate requires the full outcome field and an assumption about allocating its overround.
Compare the actual sportsbook payout with the exchange ask, fees and available quantity. A normalized no-vig benchmark can help you compare prices, but it does not change either offered payout. The odds converter guide works through those distinctions.
Bids, asks and depth
An order book contains resting orders. A 15c bid and 16c ask mean someone offers to buy at 15c and someone offers to sell at 16c, subject to the displayed quantity. A limit bid may remain unfilled; taking an ask may require several prices if your quantity exceeds the best level.
The spread is one cost indicator. It does not include all exchange fees, funding costs, price movement or market impact. Check the applicable current rules and fee schedule at the actual venue. Polymarket and Kalshi do not have a universal zero-fee or fixed 2c cost that applies to every market.
You may be able to sell before settlement if another participant will buy. That does not guarantee an exit, a particular price or a profit. Some sportsbooks offer cash-out options too, under their own terms.
What prices can tell you
Trading prices aggregate participants’ orders, but shared information, thin liquidity, risk limits and stale quotes can affect them. This guide does not establish that a prediction market is always more accurate than a sportsbook or within a fixed number of points of the true probability.
To test forecasting quality, record predictions before the relevant outcome information arrives, define the population and inspect reliability-bin counts and uncertainty. A last quote near settlement can already contain outcome information. ZenHodl’s dated regular-season report is one scoped model evaluation, not a guarantee for every current sport or trade.
Expected value is conditional on the probability
For a hypothetical contract costing 40c with a $1 winning payout and a true win probability of 50%, gross expected profit is 0.50 - 0.40 = 10c per share, or 25% of the acquisition cost. At a true probability of 45%, it is 5c before costs.
A model estimate of 50% is not proof that the true chance is 50%. Finite outcomes vary, and selected trading populations can have different calibration from the overall model. Include non-fills and costs when evaluating a strategy rather than treating the example as a promise of profit after hundreds of trades.
Rules and access come before funding
Check the venue’s current eligibility, identity checks, funding options, withdrawal terms and market rules directly on Kalshi, Polymarket or Manifold. Platform availability and legal restrictions depend on location and can change; this introductory article is not a current jurisdiction or onboarding guide.
For practice, inspect the actual rules of any paper or play-money environment. Its fills, incentives and costs may differ from real-money markets.
Size is a separate decision
Kelly sizing computes a fraction under specified probabilities and payoffs. It neither validates a model nor makes a position safe. Fractional Kelly reduces exposure under the same assumptions, but is not a universal recommended percentage; correlated positions and uncertain probabilities require additional analysis. See the formula and limits.
For an order, check depth on the side you will consume and calculate the weighted acquisition price for your quantity. An aggregate book total does not certify that your intended size will fill near the top quote.
Using ZenHodl’s evidence
Our samples and catalog let you inspect recorded fields and product coverage. Model-versus-price differences in Edge Finder are research candidates. They need timestamp, settlement, execution and probability checks before being treated as edge.
Start with a narrow question: which contract, at which time, at which executable price, with which costs and outcome definition? That makes a historical test inspectable and avoids mistaking attractive payout arithmetic for a demonstrated strategy.