Methodology

• Missing is not zero. Unreported statistics remain null, appear as unavailable and are excluded from eligible denominators.
• Minutes are provider-reported. Historical starting patterns are not confirmed future lineups or an independent minutes forecast.
• Samples use completed matches before the research cutoff. Retrospective records are not a historical archive of what was knowable beforehand.
• Hit rates describe observed samples. The eligible count, active filters and uncertainty matter more than an attractive percentage.
• API-Football coverage varies by competition, season, fixture and statistic. Every offered current league is eligible; historical depth fills progressively.
• Reported absences relate to specific fixtures, and may change. No report is not proof of availability.
• Same-game parlay assumptions are checked explicitly. We do not multiply correlated legs or invent a joint probability.
• A historical hit rate is not a calibrated probability or proof of betting value. Market comparisons require a real source, price and timestamp.

The legacy raw-data store is isolated rather than mixed into new samples. Accounts, billing and saved research are preserved; old identity links are not guessed into the new provider's ID space. Editorial news, weather and independent cross-provider datasets are not supplied under the single-provider policy.

Goals

Team totals use the provider's regulation score, excluding extra time and shoot-outs. Player and other box scores can include extra time: apply the extra-time exclusion when researching a 90-minute market.

xG

Expected goals as reported by API-Football where available. Model methodology is not independently verified; missing observations remain unavailable.

xGA

Expected goals conceded: the opponent's reported xG in the same match.

Shots

Total shot attempts as reported by API-Football. Check extra-time inclusion and the bookmaker's settlement rules.

SOT

Shots on target: attempts that would have entered the goal without a save or goal-line block. Woodwork is not on target.

Corners

Corners won by the team; match corners = both teams. Provider totals cover the whole match — for cup ties that reach extra time, check whether your book settles corners on 90 minutes only.

Yellow cards

Provider-reported yellow cards. Second-yellow/red-card treatment and booking-points rules may differ by bookmaker. Missing card counts are not recorded zeros.

Fouls

Provider-reported fouls committed. Unreported observations remain unavailable; no free-kick or offside statistic is substituted.

Reported minutes

API-Football's reported player minutes. Substitution times and stoppage-time adjustments are not inferred. Unknown observations appear as unavailable, never an invented zero.

Hit rate

Hits ÷ eligible games. Pushes (result exactly on an integer line) and games with missing data are excluded from the denominator and shown separately. A hit rate describes the past sample; it is not the probability of the next bet.

95% interval

Wilson score interval for the hit rate. With 8 games, a 6/8 (75%) record is consistent with anything from ~41% to ~93%.

Break-even

1 ÷ decimal odds. The win rate needed to break even at that price. Includes the bookmaker's margin.

No-vig probability

Implied probabilities rescaled to sum to 100% across ALL outcomes of the market (proportional method). Only computed when every outcome's price is present.

Overround

Sum of implied probabilities minus 100%: the bookmaker's built-in margin for that market.

Expected value

p × (decimal − 1) − (1 − p), per 1 unit staked, where p is YOUR probability or a validated model's. This app never computes EV from a raw historical hit rate.

Per 90

Total divided by reported minutes, multiplied by 90. Uses only observations with a usable denominator; a small playing-time sample can be misleading.

Venue split

Home team's home matches plus away team's away matches. Each side must meet the minimum sample on its own.

Coverage grade

A–D summary of what evidence exists BEFORE kickoff: prior team-stat matches (≥5 = 2 pts), prior player-log matches (≥5 = 2 pts), any odds (1), referee assignment (1), availability feed (1), baseline model (1). A ≥6, B ≥4, C ≥2, D <2.

Predictive model

Unavailable for the migrated dataset until chronological out-of-sample validation is completed. Historical frequencies are not a proven predictive advantage.

Expected XI (derived)

The players who started most often in the team's last five matches with lineup data. A derived estimate — not a predicted lineup from any reporter, and not confirmed.

Rest days

Days between this team's previous match (any competition in our data) and this kickoff. Matches in competitions we don't ingest are missing, so rest may be overstated.