The only independent measurement of who settles a market.
A prediction market is a promise about a future fact and a procedure for deciding who was right. The odds are public everywhere. The procedure is not measured anywhere — until a contract is disputed, and by then the position is already on. We read the rules text of every market on both major venues, score how ambiguously it is written, and hold the settlement record behind it: 132,154 markets, 758 disputed at least once, 340 twice or more, $5.4bn of volume on contested contracts.
Three kinds of buyer have a use for that and no way to build it: venues that want to know which of their own contracts are drafted badly, market makers pricing settlement risk into a spread, and terminals that need a trust flag beside an odds row. It is an awkward sale — one of those buyers is being measured — and it is a real one.
132,154 contracts, scored on five flags. Three of the five fire on what the rules do not say.
Every flag returns the clause that raised it, from the same matcher that scored it — so a quoted line cannot disagree with the number beside it.
When the matcher changed, 4.2% of the corpus moved and the before/after was published. Nothing is re-scored quietly.
- Resolution-trust scores across 132,154 markets, with clause-level evidence
- Oracle governance — who settles disputes, with what stake, how often slashed
- The dispute feed, as contracts are contested
- Settlement latency by venue and category, measured on when a market actually settled
- The methodology, the version history and the failure cases, in full
The floor is real: below $2,000 a month it is not worth the support it needs. Above it, the number depends on refresh rate, coverage and whether you want the scores or the evidence behind them.
Choose “data licensing” as the topic. No qualification form, no discovery call before a price — the price is above.
What is not for sale
Placement. A venue cannot pay to move up the trust scores, and there is no arrangement under which it could. This is the most monetisable measurement here and the most corruptible; the day it can be bought it is worth nothing to anyone, including the venue that bought it.
Exclusivity over a finding. If the data shows something about a venue, that finding is publishable and will be published. A licence buys access and support, not silence.
A number we have not measured. Coverage is what is on disk, and the gaps are documented rather than filled with estimates — the UMA voter concentration leg is the current one, and it is named as outstanding rather than implied as included.
Before a conversation
Everything below the API is public and worth an hour before an email: the 77 distinct analytical charts, the version history of every metric, and the replication study that failed and deleted a tier of this very ladder. If the method does not survive that reading, the licence is not worth buying and it is cheaper for both of us to find out now.
The consumer plans are on the pricing page and are not related to this one. Nothing here is investment advice; this is a research and measurement licence.