How We Rate Prediction Markets: ActionTrust Rating Explained

Check out the key criteria we consider when reviewing prediction markets:

Note: Payout Reliability and Fund Segregation & Safety together total 20% — close to the original combined 25% in the base framework — with the fund-safety component now fully protected as its own dimension rather than absorbed into a single compressed line.

This category is the prediction market equivalent of Odds Quality in sportsbooks. Where sportsbook quality is about vig, prediction market quality is about whether you can actually execute at the price you see, and how much that price degrades when you try to get meaningful size.

Benchmark: a well-functioning prediction market on a high-volume contract should have spreads of 1–2 cents on a $1 binary. Spreads of 5–10 cents represent meaningful friction.

This criteria has no equivalent elsewhere in the ActionTrust framework for casinos and sportsbooks.

A casino cannot misresolve a slot spin. A sportsbook has a clear event outcome to reference.

Prediction markets — especially political or event-based markets — can have genuinely contested resolutions where the platform's interpretation of contract language determines who gets paid. This has happened publicly, and it is the clearest trust signal in the vertical.

A process that technically exists but never produces reversals should score identically to no process. The test is whether it works in practice, not whether it exists on paper.

Bonus: Zero documented misresolutions in the trailing 24 months AND an independent resolution mechanism (third-party oracle or independent review panel) with at least one documented reversal on record → 1.1x to Resolution Accuracy sub-criterion.

Prediction markets exist in a regulatory grey zone that is actively being contested. The question is not only 'are you licensed' — it is 'what is your current legal exposure and is it changing.' This dimension requires more frequent review than any other in the framework.

Context: The CFTC asserts exclusive jurisdiction over event contracts. In April 2026 the Third Circuit affirmed an injunction protecting a registered exchange's sports event contracts from New Jersey enforcement, while other courts reached different conclusions. The legal landscape is unsettled and state-specific, so our prediction-market scores note jurisdictional risk explicitly and are reviewed on an accelerated cadence.

Same structural logic as sportsbooks and DFS, with the fund-safety component now separated into its own dimension (see below). This dimension covers withdrawal speed, method variety, and complaint rate.

Why this is a standalone dimension: State-licensed sportsbooks are legally required to hold player funds in segregated accounts. That requirement does not exist uniformly across prediction market platforms, particularly crypto-native ones operating outside CFTC oversight. A platform holding user funds commingled with operational capital creates a real insolvency risk that has no equivalent in licensed sportsbooks. This must be scored and disclosed.

A platform with 50 deep, liquid markets is more useful to serious traders than one with 5,000 thin markets. The scoring reflects genuine depth within covered categories, not headline market counts.

These dimensions use the same sub-criteria structure and scoring bands as the sportsbook and DFS frameworks. Bonus & Promo Transparency is weighted at 3% for prediction markets, reflecting that promotional activity is a smaller part of the prediction markets user experience compared to sportsbooks or DFS. See the general ActionTrust overview page for the full methodology on these shared dimensions.

The example below shows a hypothetical CFTC-registered prediction market that performs well on regulation and market quality, but receives a penalty in Resolution Accuracy after a documented misresolution was not corrected following appeal.

Final ActionTrust Score: 7.2 / 10. The documented misresolution reduced the Resolution Accuracy dimension from 7.5 to 3.75 after the 0.5x penalty, lowering the overall score by 0.75 points. Without this penalty, the platform would have scored 7.9 / 10.