Methodology

How we grade claims

We don't take sides. We test narratives against the data and show our work.

How we pick claims

Widely shared. We prioritize claims with real reach (high-engagement X posts, mainstream coverage), not obscure accounts.
Testable. The claim must make a specific statement that public data could confirm or refute: a record, a streak, a stat, a line move, a cause. Opinions and predictions about the future aren't graded until they can be.
Claims, not people. We quote the claim and link its source so readers can see the context, but we grade the statement, not the author. We don't keep scorecards on individuals.

Evidence hierarchy

Primary sources: official results and box scores, league or team injury reports, official statistics, and the published lines or odds themselves.
Reputable secondary sources: established stats sites and news organizations that report primary data, cited by name with a link.
Analysis and commentary: useful for context and for spotting disagreement between sources, never enough on its own.
Social posts are claims, not proof. A post is what we test. It's never the evidence for itself or for another post.

When credible sources disagree (for example, two books with different closing lines), we say so. That disagreement often decides the verdict.

Verdicts

Supported: The best available public data is consistent with the claim as stated.
Contradicted: The best available public data shows the claim, as stated, is false.
Mixed: Part of the claim holds and part doesn't, or it holds only under one reasonable definition or data source.
Insufficient evidence: Public data can't settle the claim either way, or the claim isn't testable as stated.

We never fill a gap with a guess. If we can't verify a number, the verdict is Insufficient evidence.

Confidence vs. evidence quality

The verdict says what the evidence shows. Confidence (High, Medium or Low) says how solid that evidence is: the quality of the sources, whether independent sources agree, sample size, and how sensitive the result is to definitions such as which line counts as the close. A claim can be Supported with only Medium confidence when the sources are thin or one data point is disputed.

Timestamps

Sports data changes by the hour. Every check lists when each data source was pulled, in US Central time, and the publish time of each cited article. A verdict is about the data as of that time. Later games or line moves can change it, and if they do, we update the check and log the change.

What we don't publish

No picks, tips or betting recommendations. We grade claims; we don't tell anyone what to bet.
No prices from data sources without permission. We don't republish prices or data from a provider (for example, prediction-market or sportsbook data feeds) unless its terms allow it or we have written permission. Where we can't show a number, we cite where it's publicly reported.
No raw data redistribution. We publish derived findings and short factual excerpts, not copies of anyone's dataset.

Mistakes

We'll get things wrong sometimes. When we do, we fix them in the open. See the corrections policy.

AI-assisted research

We use AI tools to help find claims, gather sources and draft write-ups. A human reviews and approves everything before it's published. More on our About page.