Menu
Open app
Methodology

The methodology starts with what is actually connected.

HAPPENSNEXT does not present broad domain coverage as a finished forecasting engine. The first grounded contract is Bundesliga 1X2 and it is explicitly experimental.

01

Evidence is point-in-time bounded.

The football feature bundle uses only completed match results available before the forecast cutoff. Market odds are not used by the current Poisson predictor.

Completed pre-cutoff matches only.
Provider availability timestamps.
No market odds in the current predictor.
02

Probability is not calibration.

The first football route returns a valid numerical vector but does not claim calibration. Data sufficiency, quality and freshness replace an invented confidence score.

Uncalibrated experimental label.
Structured evidence quality.
Abstain when data is insufficient.
03

Outcomes close the loop.

When the final score is available from the registered provider, the resolver appends a resolution and score. The old ECB benchmark lineage remains invalidated and read-only.

Record before kickoff.
Resolve only against the final score.
No public accuracy superiority claim yet.

Governed probability

For the supported route, the registered specialist predictor owns the stored 1X2 vector.

Evidence quality

The card reports pre-cutoff inputs, source timestamps, data sufficiency and limitations.

Immutable record

Updates are new revisions. The original probability stays preserved until reality resolves it.