How to read MarketEdgeFootball predictions
This guide explains the main labels and warning flags used by Jarvis. The key rule is simple: a strong model pick is not automatically a bet. Always check data freshness, price, lineup status and volatility before acting.
Decision labels
Use these labels as the first filter before looking deeper at the match card.
The model-side betting direction before the final price check. It is based on the strongest probability after flags and volatility are considered.
Action: Use as a shortlist item only. Check odds, data freshness and lineups before staking.
The strongest available betting candidate. Odds-backed Best Bets rank above model-only Projected Bets because the price has been checked against Jarvis probability.
Action: Can be considered first, but still avoid if data becomes stale or lineups create a new warning.
The model likes the pick and the data quality is acceptable.
Action: Can be considered, but still check the latest price and lineups.
There is something interesting, but not enough confirmation yet.
Action: Do not rush. Recheck closer to kickoff or wait for better odds.
Too much uncertainty, weak edge, stale data or high volatility.
Action: Skip the market unless new information changes the picture.
The pick may be good only at a minimum price.
Action: Only enter if the available odds are equal to or better than the target price.
The match is sensitive to starting XI, rotation or late availability.
Action: Wait until confirmed lineups before taking a serious position.
Critical data is old or missing.
Action: Do not rely on the current prediction until the data refreshes.
What the risk flags mean
A flag does not always mean the opposite bet is correct. It means the match needs extra caution.
A draw-specific betting lean. It appears when the draw probability is high enough, usually together with Draw Watch or a compressed 1X2 market shape.
Action: Do not treat it as a banker. Use only if the draw price is attractive, and consider smaller stake or double-chance style cover where available.
The model sees elevated draw risk, usually from balanced teams, low scoring profile or cautious incentives.
Action: Avoid aggressive single-team bets. Consider draw, double chance, or no bet depending on price.
A favourite looks attractive on paper, but volatility, market price, motivation or matchup risk makes it dangerous.
Action: Do not blindly back the favourite. Reduce stake, wait for lineups, or consider draw/underdog cover.
The underdog has a realistic chance to outperform market expectation.
Action: Look at underdog double chance, draw-no-bet, handicap, or avoid favourite-heavy accumulators.
Goalkeeper quality, availability, errors, penalty profile or shot-stopping may materially affect the match.
Action: Be careful with clean-sheet assumptions and tight favourites. Recheck confirmed lineups.
Corners, aerial strength, free kicks or defensive set-piece weakness could swing the result.
Action: Respect underdogs with strong set pieces. Be careful in low-margin matches.
The expected XI and actual XI may differ materially, or rotation risk is high.
Action: Wait for confirmed lineups. Re-run or recheck the match after team news is confirmed.
Sources disagree or the news is not yet confirmed.
Action: Downgrade confidence. Do not act until the conflict is resolved by a reliable source.
Jarvis and the betting market strongly disagree.
Action: Treat as opportunity only if data quality is high; otherwise reduce confidence or avoid.
The match has a wider range of possible outcomes.
Action: Avoid large stakes and avoid using the match as a banker in accumulators.
Internal model layers do not agree strongly.
Action: Use caution. The final pick may be fragile or scenario-dependent.
Core prediction terms
These appear across match cards, watchlists and value screens.
The model’s single preferred outcome or betting direction after combining rating, squad, market and risk layers.
How strongly the model supports the pick. Confidence is not the same as betting value.
How unstable the match is expected to be. High volatility means more upset/draw risk.
How fresh and complete the important inputs are: fixtures, squads, news, availability, odds and simulation data.
Difference between Jarvis probability and the market-implied probability.
The price implied by the model probability. If the market price is higher, there may be value.
Matches that pass stricter filters for confidence, data quality and value discipline.
Safe, Balanced and Risk-on accumulators are not guarantees; they are ranked by risk and correlation.
Before taking action
The model should be treated differently depending on freshness and confirmation status.
Data appears fresh. You can consider the model output, subject to price and lineup checks.
Some data is degraded or the match is sensitive. Wait for a refresh or check manually.
Critical information is stale or missing. Do not rely on that prediction for betting.
How to use the site in 60 seconds
A practical order for checking matches without overcomplicating the decision.
If stale or blocked, do not bet blindly.
Look at confidence and volatility together.
Draw Watch, Trap and Upset Flag change the staking decision.
Only act if the odds still make sense and lineups do not damage the pick.