Original research
The sources behind 103,998 Kalshi contracts
ESPN appeared in the settlement-source lists for 40.2% of the open contracts in our snapshot. We mapped the publishers, leagues and data providers named.
Prediction Contracts editorial · · 4 min read
ESPN appeared in the settlement-source lists for 41,787 of the 103,998 open Kalshi contracts we collected on September 6. That is 40.2% of the sample. Restrict the count to contracts traded during the preceding 24 hours, and the share rises to 43.5%: 8,144 out of 18,721.
Those figures describe where ESPN was named as a source, not which results it decided. Many contracts list more than one source. But they reveal a part of prediction markets that receives less attention than the odds: the evidence used to determine a payout.
The price and the result come from different processes
Traders buy and sell claims about an event. The contract’s rules define what counts as the outcome and where the relevant evidence comes from. A football statistic might come from a league or a sports publisher. A price threshold might use a benchmark feed. Knowing the market’s forecast does not tell you which of those sources governs its result.
Kalshi’s series API exposes settlement-source lists alongside links to contract terms. We joined those lists to open contracts and counted each source hostname once per contract. A series is a family of related contracts; an event can contain several contracts, such as different thresholds or candidates. Our downloadable source map separates contract counts from distinct event counts.
| Source hostname | Open contracts | Distinct events | Contracts traded in preceding 24h |
|---|---|---|---|
| espn.com | 41,787 | 4,584 | 8,144 |
| foxsports.com | 17,721 | 2,583 | 3,721 |
| nfl.com | 13,668 | 1,071 | 2,643 |
| ncaa.com | 12,854 | 1,173 | 2,555 |
| apnews.com | 8,652 | 846 | 1,206 |
| wsj.com | 4,508 | 727 | 1,239 |
| reuters.com | 4,412 | 691 | 947 |
| nytimes.com | 4,322 | 746 | 1,024 |
ESPN’s reach is not just a count of individual contracts
The ESPN references spanned 4,584 of the 12,645 event tickers in the sample, or 36.3%. Within the Sports category, ESPN appeared for 41,630 of 62,410 contracts: 66.7%. The result is therefore substantial whether counted by contracts or by events, although neither measure tells us how much money or how many people depended on that source.
Sports made up about 60% of this open-contract sample. We collected it late Saturday, September 5, in New York—September 6 in UTC—during football season, and half the contracts reported no lifetime trading volume. Those details matter when interpreting the 40.2% headline. It is a description of this snapshot, not an annual estimate of ESPN’s role across the exchange.
A source list is the beginning of the rules check
The receiving-yards series KXNFLSEASONRECYDS in our archive lists ESPN before the governing league, with nfl.com as the league’s URL. At our fact-check, the linked football-stat contract terms were titled FOOTBALLENTITYSTAT, despite the PDF filename FOOTBALLSEASONSTAT.pdf. Those terms explicitly put the governing league first and ESPN second in the source hierarchy. The order of names in API metadata therefore should not be treated as the contractual order of priority.
Outside sports, the map includes news publishers, public institutions and pricing services. The two copper contracts examined in our last-trade audit share a series naming Pyth’s copper feed. That is a different source arrangement from a box score or a news report, even though all three can supply evidence for a contract result.
This is useful for a reader comparing contracts: follow the source link, then read the operative terms. Similar headlines can refer to different data, release times or definitions. The settlement guide explains how those differences affect what a contract actually asks.
What we counted—and what we did not
We requested open, non-combination contracts through Kalshi’s public API and followed pagination until it returned an empty cursor. The 104 responses contained 103,998 unique contracts. Collection ran from 03:32:50 to 03:34:05 UTC on September 6, 2026. It was a sequential sweep, so listings could open or close while it ran.
We separately retrieved 13,839 unique series records, inferred each contract’s series from its event ticker and matched 103,836 contracts to a source list. The remaining 162 stayed unmatched. They remain in the overall denominator and contribute no source references; that is missing metadata in our join, not proof that those contracts lack settlement rules.
Hostnames were lowercased and a leading “www.” removed, then deduplicated within each contract. We did not merge domains into corporate parents. ESPN and ABC News, for example, remain separate entries. Each source’s recently traded count uses contracts with positive reported volume during the preceding 24 hours. That volume is a contract count, not a count of traders.
The map does not rank providers by reliability, establish exclusive dependence, or measure financial exposure. Source rows overlap and must not be added together. We checked examples of the underlying terms rather than reading every contract PDF, and did not observe which source was actually used at settlement. The downloadable methods explain the definitions and limitations.
The contribution is a dated map of named sources, not a claim to have discovered that markets need outside evidence. It gives readers a way to see which organizations are written into the contracts—and a starting point for checking the rules behind a quoted probability.