Who We Are, How the Model Works & What We Stand For

xG is the foundation of Over/Under, Correct Score and BTTS tips. OddVora uses context-adjusted xG averages: home xG and away xG are calculated separately, because a team's home xG is typically 30–40% higher than their away xG. Averages cover the last 12–15 fixtures and update within 24 hours. For Correct Score, xG drives a Poisson distribution generating per-scoreline probabilities. For BTTS, both teams' xG-against averages calibrate mutual scoring probability.

Results over the last 8–12 games are weighted by the Elo-adjusted quality of each opponent. A win against a top-six side carries three times the weight of a win against a relegation candidate. This prevents the model over-rating a team on a strong run against weak opposition. Form is secondary — good form combined with poor xG metrics does not produce a strong tip.

Applied three ways: as a BTTS frequency signal (75%+ mutual scoring in last 8 H2H elevates the BTTS prior); as a correct score repetition premium (an exact scoreline in 2+ of the last 4 H2H meetings receives a +25% Poisson premium); and as a dominance signal for 1X2 (80%+ H2H win rate adjusts the match result probability). All H2H data is venue-filtered — home records only for home tip calibration.

Market prices are not a primary model input — that would create a circular system. Instead they serve as a calibration check (a probability wildly different from all bookmakers flags a potential data error) and as an edge detector for Hot Picks and Long Shots (the edge figure per tip is model probability minus bookmaker implied probability).

Applied after the quantitative model produces its base probabilities. Includes fixture congestion, confirmed rotation signals from press conferences, motivational differentials and European travel fatigue. These carry low weight — typically 3–6 percentage point adjustments. The Banker of the Day qualification requires a situational review as its mandatory final step before publication.

Shot-level xG data including position, body part, shot type and context, processed through an internal Poisson model calibrated against eight seasons of outcome data. Updated within 24 hours of each completed fixture. UCL xG is sourced from UEFA's official data partner.

Official final scores, goalscorers, half-time scores, clean sheets and disciplinary records are processed within two hours of the final whistle. The H2H database holds eight seasons of records per European competition, calculated separately by venue. NPFL records cover four seasons.

Confirmed absences from official club sources and press conference reports are reviewed before tips are finalised. The Banker of the Day explicitly requires a team news check — a material absence (goalkeeper, both first-choice centre-backs, or a forward accounting for 35%+ of competition goals) suspends the banker designation regardless of statistical confidence.

xG data is not available for NPFL fixtures. The model uses scoring frequency rates, clean sheet percentages and H2H records as proxies. This produces higher prediction uncertainty. NPFL confidence scores are typically lower and NPFL tips carry explicit variance disclosures on every market page.

Every tip is logged with a timestamp before kick-off. Once published it cannot be deleted, hidden or reclassified. The full archive — including the November 2024 loss cluster — is preserved in the results and track record pages.

Results are updated from official verified scores after the final whistle. No result has been reclassified as a void, changed from a loss to a win, or modified for any reason since launch. The record you see is the record that happened.

November 2024 (−0.6% yield, our only losing month) appears on the track record page in the same format as February 2026 (+22.6%, our best). It includes a full post-mortem: which markets failed, why, and what model changes followed. Selective presentation of results is the primary mechanism dishonest tipsters use to inflate apparent performance.

OddVora earns revenue through advertising and regulated bookmaker affiliate partnerships. These are disclosed where affiliate links appear on the odds comparison page. They generate revenue when users sign up — not when they win or lose. Commercial relationships do not influence tip selection, model calibration or results reporting.

No subscription, no registration, no payment of any kind. This is the permanent structure of OddVora, not a promotional policy. See the 19-month track record and judge for yourself whether free tips can be genuinely good tips.

Even the highest-confidence selection carries a real probability of loss. The Banker of the Day loses roughly 1 in 11 tips. Long shots lose roughly 8 in 10. Your staking plan must account for this.

Every regulated bookmaker offers deposit, loss and session time limits. Set these before your first bet — not after a losing run. Proactive limits protect you from variance in a way that reactive limits cannot.

Increasing stakes after a losing run to recover losses is the single most dangerous behaviour in sports betting. A run of 5–10 consecutive losses is statistically normal in every market we cover. Chasing amplifies variance — it does not reduce it.

A single week of results tells you almost nothing about underlying model edge. OddVora's +16.8% all-time yield is meaningful because it covers 4,218 tips over 19 months. Assess any betting strategy over at least 30 settled tips before drawing conclusions.