Editorial standards
PUBLISHING POLICYThe rules we use to describe Gary's authorship, research, verification, confidence, grading, and mistakes.
State the authorship plainly
Gary is an AI sports-analysis product and editorial persona. A Gary byline or first-person sentence is the product voice, not evidence that a human handicapper wrote or reviewed the analysis. We only describe human review on a page if it actually occurred and is identified there.
Research before the call
The workflow begins with an AI research agent that gathers available market, schedule, statistical, availability, matchup, venue, and weather context. Not every category applies to every sport or game, and a missing input should not be filled with an invented fact.
Check factual numbers
A separate pass checks numeric claims in the rationale against the data fetched for that analysis. A claim that does not survive that check is corrected or retried rather than knowingly published as verified.
Keep the sport-specific rules consistent
Gary applies a written set of considerations for each sport instead of treating every market alike. The weight of starting pitching, guard play, rest, venue, or other context depends on the sport and matchup.
Show uncertainty
Confidence is an estimate attached to the analysis, not a guarantee. When no call has posted, the board says that it is pending or that no call was posted instead of presenting a placeholder as a recommendation.
Grade the outcome in public
Published picks are graded against final scores as wins, losses, or pushes. Win percentage uses decided results, with pushes excluded, and unit results use flat one-unit stakes at the listed odds. Losing results remain part of the record.
What the process cannot promise
Data can be delayed, incomplete, or later revised. Sportsbook prices can move after research. AI-generated analysis can still contain an error, and a well-supported call can still lose. Gary's content is informational and entertainment content, not a promise of profit or a substitute for checking current information yourself.
See the full methodology, learn which data categories inform the work, inspect the public record, or report a factual error.