When there is no data feed, the broadcast becomes the data.
PLLAY Vision analyzes authorized recorded video; live reads are human-confirmed; unclear evidence is held or refunded (fail-closed).
From video to verified outcome.
One recorded run, read end to end: the frame PLLAY Vision sees, the evidence it extracts from it, and the outcome record it produces. Every figure below is read from that run, not typed.
Illustrative interface. Commercial deployments are configured for approved content, event definitions and evidence requirements.
Capture. Interpret. Reconcile. Resolve.
Five steps, in the order every deployment runs them. PLLAY Vision works on authorized recorded video and hands its outcome to downstream systems; it does not settle live broadcasts on its own.
- 01Capture
Ingest authorized recorded video from the approved content workflow.
Produces · Frames - 02Interpret
Identify relevant gameplay signals, objects, actions and state changes.
Produces · Signals - 03Reconcile
Compare multiple frames and evidence signals instead of relying on a single visual event.
Produces · Reconciled reads - 04Score
Calculate outcome confidence against the configured event rules.
Produces · Confidence - 05Resolve
Generate an evidence-backed outcome record for downstream engagement and reward systems.
Produces · Outcome record
The same five steps run for every deployment; only the approved content, event definitions and evidence requirements change.
Validated per title.
- Apex Legends
- Call of Duty: Warzone
- Counter-Strike 2
- Dota 2
- Fortnite
- League of Legends
- Madden NFL 26
- NBA 2K26
- Overwatch 2
- PUBG: Battlegrounds
- Rainbow Six Siege
- Rocket League
- Rocket League Sideswipe
- Valorant
Adapters are validated per title against the published benchmark record. A title with no validated adapter is not a supported one. See all titles
Technical detail and validation record
Every deployment is configured for approved content, event definitions and evidence requirements before it produces an outcome, and an outcome below the configured confidence bar is held for review rather than asserted. The full validation record, per title and per confidence tier, is published at the benchmark page; the fan-facing reading of the same system is AI Referee.
Built for evidence, not just detection.
Detection finds a moment. Evidence is what lets an outcome carry a prediction, a challenge or a reward.
Multi-signal reasoning
PLLAY evaluates sequences of evidence rather than treating one frame as the truth.
In the run above: the score bar across the cut, then the final banner, compared before anything is called.
Configurable event logic
Each deployment uses approved event definitions, confidence thresholds and resolution rules.
The rule that settled the run above — terminal, stable, unanimous — is one configuration, not a fixed behaviour.
Auditable outcomes
Every resolved event can retain its supporting evidence, timestamps and decision record.
The record above is the run log itself: every read, its confidence and its timestamp.
Human review controls
Low-confidence or disputed outcomes can be routed into a controlled review workflow.
Below the configured confidence bar, an outcome is held for review instead of resolved.
One outcome layer. Multiple commercial use cases.
Wherever a result has to be trusted before it pays out.
- Sponsored predictions
Brand-funded predictions around recorded gaming content.
- Creator challenges
Creator challenges with evidence-backed results.
- Reward campaigns
Reward campaigns that require verifiable completion.
- Platforms without data
Platforms without reliable structured gameplay data.
PLLAY Vision can operate alongside PLLAY APIs, embeds and overlays or as an outcome-intelligence component within an approved partner workflow.
Turn video into actionable outcomes.
Deploy PLLAY Vision for an authorized recorded-video workflow with configured event logic, evidence standards, outcome records and integration support.