Prediction market pricing is the process of turning a live probability estimate into a tradable price, usually a contract worth $1 if an outcome happens and $0 if it doesn’t. Simple enough. Except nobody “sets” that price the way a bookmaker sets odds, and the number on your screen is the product of two very different systems shaking hands: an order book full of traders, and a data pipeline carrying play-by-play information out of a stadium in a fraction of a second.
This guide walks through that pipeline in order, from the field to the price, and uses Genius Sports’ new prediction market venture as a concrete illustration of how the data-to-pricing workflow is being commercialised.
What is prediction market pricing?
A prediction market is a venue where people trade contracts on whether a specific event will happen. Sports contracts are typically binary: “Team A wins this game” settles at $1 (or 100 cents) if true, $0 if not. Because the payout is fixed, the price itself reads as a probability. A contract trading at 62 cents implies roughly a 62% chance, the equivalent of decimal odds around 1.61.
That’s the distinction worth holding onto. A sportsbook quotes a price it is willing to take the other side of, with a margin baked in, the overround. Across a two-way market, the bookmaker’s implied probabilities sum to more than 100%, and that gap is its edge. In a prediction market, price discovery happens between participants. The venue earns from trading fees, spreads and, in some models, its own market-making activity rather than from a posted margin.
Now the complication: “no margin” does not mean no cost. The bid-ask spread is a real cost. If the market on a first-half total shows 60 bid and 64 offered, you’re buying at 64 and could only sell immediately at 60. Thin markets widen that gap. Add commissions, withdrawal frictions and the risk of getting filled at a stale price during a fast-moving sequence, and event contracts carry the same structural reality as any other wagering product: most participants lose money over time. Regulatory treatment also differs by venue and jurisdiction, with US event contracts falling under CFTC oversight rather than state gaming regulators.
The live data pipeline: from field to market
Every in-play price depends on knowing the current state of the game before the crowd does. Here’s the sequence, in the order it actually happens.
Step 1: data collection
Sports data leaves the venue through a handful of channels, usually in combination:
- In-venue human collectors. Trained operators at the ground log events, possessions, shots, fouls, substitutions, into a tablet application. Official data rights deals exist largely so this happens from an accredited seat rather than from a TV screen at a delay.
- Optical and computer vision tracking. Fixed cameras track player and ball positions many times per second, producing spatial data that human collection can’t capture.
- Sensors and chips. Wearables and ball-embedded devices in some leagues add velocity, distance and location detail.
- League and official scoring systems. The scoreboard, clock and official stat feed act as the reference truth for settlement.
The commercial reason data providers pay leagues for exclusivity is straightforward: the fastest, most authoritative version of an event is worth more than a slower copy of it. Scraped or broadcast-derived data arrives seconds later, and in an in-play market, seconds are an eternity.
Step 2: processing and validation
Raw collection is noisy. A collector mis-taps, a camera loses a player in a ruck, a feed drops packets. So the data passes through a validation layer before it ever touches a pricing engine. Typical checks include cross-referencing two independent sources, sanity-testing against game rules (a basketball possession can’t produce five points), flagging implausible clock jumps, and reconciling against the official feed.
When something fails validation, the usual response is not to guess. Markets get suspended. You’ve seen this as a bettor: the “market unavailable” moment right after a goal-mouth scramble. That pause is the system protecting itself while it confirms what happened.
Step 3: price calculation
Validated state data feeds a probability model, which outputs a fair value for every open contract. The model takes the current score, time remaining, possession, personnel and pre-match expectations, and asks: from this exact position, how often does each outcome occur?
Those models are generally built from large historical samples of comparable game states, sometimes supplemented by simulation, running a match forward thousands of times from the current position. The output is a probability, which converts straight into a price. A 41% win probability is a 41-cent contract before any spread or fee is applied.
Key data points that drive pricing
Not all inputs move a price equally. A substitution in the 20th minute barely registers; a red card in the 70th rewrites the whole book. The table below maps the main feeds to their pricing effect.
| Data point | Typical source | Why it moves the price |
|---|---|---|
| Score and game clock | Official scoring feed | The dominant input. Score plus time remaining defines the game state the model prices from. |
| Possession and field position | In-venue collection, tracking | Drives short-horizon markets: next score, drive outcome, next goal. |
| Player availability and lineups | Team announcements, official feeds | Pre-match repricing. A starting quarterback or a top scorer sitting out shifts the base probability materially. |
| In-game injuries and cards | Live collection, broadcast confirmation | Changes team strength mid-event, often the largest single in-play repricing after a score. |
| Player and team statistics | Historical and live stat feeds | Feeds player prop pricing and the priors behind team-level models. |
| Weather and pitch conditions | Meteorological feeds, venue reports | Mainly affects totals and scoring-rate markets in outdoor sports. |
| Order flow and liquidity | The exchange itself | Not sports data, but it shapes the traded price: where size sits decides whether a model’s fair value survives contact with the book. |
Genius Sports’ prediction markets play: a worked example
In September, Genius Sports launched Prediction.com, a dedicated prediction market affiliate site built by Legend, the consumer media and technology unit it acquired for $1.2 billion. The site aggregates pricing and market information from multiple prediction market operations and overlays it with live sports feeds.
It’s a useful case study because it makes the data-to-pricing workflow visible to the end user rather than hiding it inside a trading desk. Three features illustrate the point:
- Official in-game data sits alongside market probabilities. Users can track play-by-play action next to dynamic market prices, probabilities and signals, effectively watching the input and the output at the same time.
- Like-for-like contracts are compared across venues. Because equivalent contracts trade on several platforms at once, the same outcome can carry different prices. Aggregation exposes that gap.
- A proprietary cross-venue engine handles multi-leg positions. Comparing multi-leg contract combinations across exchanges is a pricing problem, not just a display problem, which is why it needs an engine behind it.
Genius framed the launch around fragmentation: users were visiting platforms separately to access pricing information. Sean Conroy, EVP of rights and partnerships, put integrity at the centre of it, saying that as the category grows, “robust integrity safeguards are not simply optional; they are foundational,” and describing the combination of official data, live streaming, league IP, integrity services and Legend’s media reach as what helps Polymarket deliver a differentiated experience for US sports fans.
The commercial backdrop matters too. Genius signed deals with both Polymarket and Kalshi in August to supply real-time event data, league integrity services and streaming, and rival data provider Sportradar announced similar multi-year partnerships with the same two venues in June. Prediction.com launched as the NFL season began, a peak period for sports data monetisation, with Genius also holding DraftKings and FanDuel partnerships. Citizens gaming analyst Jordan Bender noted that with the NFL season underway, Genius “should benefit from increased spending across both traditional sports betting and the emerging prediction market industry,” and that investors would be watching whether the company can use Legend’s relationships to capture more of that opportunity.
Read structurally, both providers are selling the same thing to prediction markets that they already sell to sportsbooks: the authoritative version of what just happened, fast enough to price on, plus the integrity monitoring that makes the resulting market defensible.
How prediction market odds are set
Model fair value is an opinion. The traded price is a negotiation. The mechanics work roughly like this:
- Market makers post two-sided quotes. Professional participants, often running their own probability models on licensed data feeds, place bids and offers around their fair value. The gap between them is the spread, and it compensates them for the risk of being wrong or being picked off.
- Informed traders take the other side. Anyone who believes the posted price is wrong trades against it. That flow is what actually moves the price.
- Liquidity sets the elasticity. In a deep NFL moneyline, a modest order barely shifts the price. In a thin player prop, the same order can move it several cents. This is why two venues can show different prices for identical outcomes.
- New information resets the whole book. A touchdown or a red card invalidates every open quote at once. Makers pull or reprice, and the market re-forms around the new game state.
Update frequency is therefore event-driven rather than on a fixed tick. Between plays, prices drift gently as time decays. On a scoring event, they gap. For genuinely instant markets, repricing happens as fast as the data arrives and the makers can react, which in practice means well under a second for the systems and slower than that for humans clicking a mouse.
Speed and accuracy: why milliseconds matter
Latency is the entire game in in-play pricing. If your data reaches you later than someone else’s, you are quoting yesterday’s weather. Three consequences follow.
Stale quotes get picked off. A trader with a faster feed knows the goal has gone in before the slower market maker has pulled its offer. That’s courtsiding logic applied to an order book, and it’s why official low-latency data is priced as a premium product rather than a commodity.
Bad data creates bad settlements. If a validation error puts the wrong score into the model, prices form around a fiction. Correcting it afterwards means voided trades, disputed settlements and reputational damage, which is far more expensive than suspending a market for ten seconds.
Integrity monitoring depends on the same feed. Unusual price movement ahead of an event on the field is one of the clearest signals of an integrity problem, but you can only spot it if your data timestamps are trustworthy. That’s why data supply and integrity services are typically sold together.
For anyone trading these markets rather than building them, the practical takeaway is uncomfortable but useful: you are almost certainly not the fastest participant in the book. Your edge, if you have one, has to come from judgement about the model, not from reacting to a broadcast you’re watching on a delay.
FAQ
How are prediction market odds calculated?
They aren’t calculated centrally. Participants and market makers derive fair value from probability models fed by live sports data, then post bids and offers. The traded price is where supply meets demand, and because contracts settle at $1 or $0, that price reads directly as an implied probability.
What data feeds prediction markets?
Official league data supplied by providers such as Genius Sports and Sportradar: scores, clock, possession, lineups, player statistics, injuries and card events, plus optical tracking in some sports and contextual feeds like weather. Both companies announced data and integrity partnerships with Kalshi and Polymarket in 2025.
How fast do prediction markets update?
Event-driven rather than on a schedule. Prices drift slowly between plays and gap immediately on a score, card or injury, with automated systems repricing as soon as validated data lands. Markets are often suspended for a few seconds while ambiguous events are confirmed.
What technology powers sports betting odds?
The same stack in both worlds: in-venue and optical data collection, a validation layer, API distribution to clients, probability and simulation models, and a pricing or trading engine. The difference is what sits at the end. A sportsbook posts a margined price; a prediction market posts an order book.
Event contracts and sports wagering both carry a real risk of losing money, and no model or data advantage removes that. Treat them as speculation rather than income, set deposit and loss limits, and use self-exclusion or cool-off tools if the activity stops feeling like a choice. This article is educational and is not financial or betting advice.