This page describes the standards gambeta.ai uses to produce and publish its predictions. It is written so that anyone —a reader, a journalist or a search engine— can understand exactly what sits behind each pick and verify it independently.
Who writes and who reviews
The predictions are generated by an artificial-intelligence model, not a person writing by hand. We say so plainly at the foot of every article: written by Gambeta AI, reviewed by Mauro. Mauro is the responsible editor: he oversees the system, checks that picks only go out with real data, and keeps the public record. We do not invent fake human authors to look like more bylines than there are.
What data goes into a pick
The model combines two sources for each match and only publishes when it finds value —that is, when its probability differs from the one implied by the bookmakers' odds—. The technical detail is in how the model works. In short:
- The market: the real bookmaker odds, from which the implied probability is extracted after removing the margin.
- The match data: recent form, goals scored and conceded, head-to-head, home advantage, injuries and confirmed line-ups and the appointed referee; in baseball the starting pitcher; in tennis, the gap in level by the tier of the tournament.
A rule we never bend: if a match lacks the data it needs, no pick is published for that match. We prefer a few well-grounded picks to one for every game. The minimum publishable odds are 1.50.
How results are settled
Every pick is recorded with its date and its odds before kick-off. The result is settled automatically against independent public sources (ESPN, TheSportsDB and official result data). When a score is in doubt, it is checked by hand against sports sources before the pick is closed.
- A postponed or suspended match is marked as voided and counts as neither a win nor a loss.
- Each pick is marked as won, lost or void, with the final score on view.
Corrections and the integrity of the record
This is the point that sets us apart from almost the whole industry:
- The record is not edited after the fact. A published pick is never deleted or rewritten. If it loses, it is recorded as a loss.
- Losses stay up alongside wins in the verified track record, with their odds and their result.
- If we find a data error (say, a result settled wrongly), we fix the system that produces it, not the record: the correction is visible, not hidden.
📊 The whole record is public and downloadable as open data (JSON and CSV, CC BY 4.0). The point of showing it is that anyone can recompute the numbers themselves.
What you will not find here
- We do not promise profits. No AI wins every time; anyone who tells you otherwise is lying. Predictions are estimates and do not guarantee results.
- We do not publish inflated hit rates or cherry-picked "streaks": the number that matters is the full, live record.
- We do not edit a pick after the match to make it look like it landed.
Independence and how we are funded
The AI predictions and the full record are free and public, with no mandatory sign-up. The project is sustained by its official sponsor and by affiliate links to bookmakers, not by charging the reader to see the AI picks. When a link is an affiliate one, it is because that relationship exists; it never changes what the model publishes. We also offer optional paid services (human picks and VIP communities) that are clearly separate from the free AI product.
Report an error
If you see a mis-recorded result, a wrong data point or anything that does not add up, write to us from the contact page or on X at @Gambeta_ai. We fix it and, where it matters, explain it.
🔞 gambeta.ai is for adults 18+ only. Betting carries risk and can be addictive: read our responsible gambling page.
More about us
- How the model works — the technical detail of the methodology.
- Mauro, editor and reviewer — the person who oversees the picks.
- Verified track record — every pick with its result, wins and losses.