01 / THE PROCESS
From football facts to a decision
AI betting analysis is a structured estimate, not a way to know tomorrow’s score. Our paid pre-match analysis combines numerical models with match context. Live Alerts use a separate process; these pre-match rules should not be read as live-Alert thresholds.
1. Identify the fixture and assess the available information
Check teams, competition, kick-off and home/away context. Recent results, attacking and defensive indicators, expected goals when available, rest and absences inform the assessment. Missing information is a limitation—not permission to invent it.
2. Ask what each team actually needs
Standings, qualification rules, aggregate scores and the next match can change incentives. An already-qualified favourite is not automatically a bet against: its opponent still needs the ability to exploit any rotation. A draw suiting both teams does not prove they will draw.
3. Estimate outcomes, then challenge the estimate
The numerical pipeline combines its result estimate with an independent model when available, then pulls the 1X2 estimate halfway toward the margin-free market. This reduces extreme disagreement; it does not make the estimate infallible. A model–market gap above 15 percentage points blocks a result-value claim.
4. Check the exact selection at the actual price
Home win, double chance, goals and Asian handicap have different settlement rules. We evaluate supported markets separately rather than assuming the most likely winner is the best-priced selection. Confidence and data quality are separate from win probability.
5. Allow abstention and limit exposure
A selection without enough supported edge can remain information-only or be passed. Suggested single-bet exposure in paid analysis never exceeds 2% of bankroll and can be lower or zero. That ceiling limits a suggestion; it does not make a bet safe.
02 / THE PRICE MATTERS
Likely to win is not the same as value
For a simple win-or-lose market, the break-even probability is 1 ÷ decimal odds. At 1.50 it is 66.67%; at 2.00 it is 50%. These are price calculations—not our predictions.
A positive figure depends on the probability estimate being reliable. Lower odds can remove the apparent edge without changing the match. The bookmaker’s margin is removed from a complete outcome market when comparing market probabilities; expected return is checked at the actual quoted price.
For paid pre-match analysis, a market must reach at least 2% estimated expected return to qualify for a value label or a suggested stake. Below 1%, a checked market is information-only and ranked last. Between 1% and under 2%, it still does not qualify for value or a stake. These are internal screening rules, not guarantees of an actual return.
Draw No Bet and Asian handicap can return all or part of the stake. They require settlement-aware calculations, not the simple binary formula above. Correlated selections in the same match also cannot be treated as independent probabilities.
We do not reconstruct a past probability from a winning score. Without a saved pre-match estimate and price, a winner cannot be called proof of positive expected value.
03 / FROM OUR RECORDED HISTORY
Real winning slips. Clear boundaries.
These public, owner-supplied slips were checked against our settled records on 5 November 2026. Each leg has a recorded final score. They illustrate how markets settle; they are not evidence that our current model generated them, and not a representative sample of all results.
Owner-supplied free slip
Recorded combined odds
3.459
Monterrey – Tijuana
3–3Both teams to score: Yes
Both teams scored, so the BTTS selection won.
Guadalajara – Atlante
4–2Over 2.5 goals
Six goals were scored; Over 2.5 required at least three.
Club América – Toluca
2–2Double chance 1X
The draw was covered by 1X. A home-win-only bet would not have won.
All listed legs settled as wins
Owner-supplied free slip
Recorded combined odds
6.048
Fiorentina – Atalanta
2–2Match result: Draw
The match finished level, settling the draw selection as a win.
Vasco da Gama – Flamengo
1–4Both teams to score: Yes
BTTS won despite a three-goal winning margin. BTTS does not require a close match.
All listed legs settled as wins
Historical odds are not available prices today. No pre-match model audit is attached to these examples, so we do not claim a probability, an edge or an original reasoning trail for them. Losses happen; selected winners cannot establish a win rate or profitability.
04 / WHAT CAN GO WRONG
The limits of AI football analysis
- Data uncertainty: a missing player is not necessarily a confirmed absence. Probable lineups must not be presented as confirmed.
- Changing conditions: a late lineup, a moved price or kick-off can invalidate an earlier assessment. Never treat a cached analysis as a live entry signal.
- Model error: historical patterns can stop working. Red cards, injuries, tactical changes and finishing variance are not fully predictable.
- Language-model limitations: readable explanations do not validate a probability. Value labels and stake controls come from numerical checks, not persuasive prose.
- Selection bias: a few winners, including the examples above, say nothing conclusive about long-term performance.
- Accumulator risk: every additional leg creates another failure point. Double chance still loses when the uncovered result occurs.
Proper evaluation needs the full dated record, actual pre-event prices, losses and voids, sample size, calibration checks and comparison with a baseline. This guide does not publish a validated long-term profit or accuracy claim.
05 / BEFORE ANY DECISION
A practical decision checklist
- Is this the right fixture, and has it started?
- Are the important absences and lineups confirmed or only reported?
- What does each team gain or lose from a draw or defeat?
- Is the exact market still available at the quoted price?
- Is the estimate supported, or is the selection only information?
- Do I understand a draw, push, half-win and half-loss for this market?
- Would I still accept the risk without seeing yesterday’s winners?
If essential facts or the price cannot be verified, abstain. No fixture requires a bet.
Questions about AI betting analysis
What is AI betting analysis?
It combines statistical estimates, football context and available market prices to evaluate a selection. An estimate is not the true probability and is not a guarantee.
Does a winning example prove the model works?
No. Selected wins demonstrate settlement only. Assessing performance requires all selections, losses and voids, a defined period, sample size and prices recorded before kick-off.
Is double chance always safer or better value?
It covers two results instead of one, but usually pays less. A higher chance of winning does not automatically make its price good value.
Does confidence mean the chance of a win?
No. Confidence reflects the quality and agreement of the analysis. It is distinct from the estimated probability of a particular market.
Can AI guarantee profit?
No. Models can be wrong, prices move, data can be incomplete and losing runs happen. Betting carries financial risk. Only participate where legal and if you are 18 or older.
Explore the analysis
Inspect a match, understand its markets and decide whether the available evidence is enough. ODDYSAI is an analysis service, not a bookmaker. 18+ · Only risk money you can afford to lose.
