๐ Today's Games
๐ Predictions lock once the match kicks off. Make sure to enter yours before the scheduled time.
๐ Predictor vs Actual โ Accuracy
๐ All Group Stage AI Predictions
๐ All Group Stage Matches
๐ Group Standings
โญ Team Rankings & Analysis
๐ Actual vs Predicted โ Full Comparison
Enter actual scores below. The predictor accuracy updates automatically.
๐ Fantasy League
All users ranked by their prediction accuracy โ including the AI Predictor.
โน๏ธ Prediction Engine
How the AI World Cup Predictor works โ every factor that goes into each score prediction.
๐งฎ Core Formula
Goal Power = (ATT ร 2.5 ร 1.5) + (MID ร 0.8) โ (opp_DEF ร 1.0) โ (opp_GK โ 6) ร 0.7 + (MID_diff ร 0.5)
Total = (GoalPower + Pedigree + Altitude + SetPiece + Cohesion) ร HomeBoost ร 0.7 + (Manager ร 0.3)
๐ Poisson Regression (2022 WC Calibrated)
| ฮป = exp(ฮฒโ + ฮฒโยทATT โ ฮฒโยทopp_DEF โ ฮฒโยทopp_GK + ฮฒโยทhome + ฮฒโ ยทformation) | Expected goals modelled as Poisson distribution. Coefficients (ฮฒ) derived from 64 matches of 2022 World Cup data, not hand-tuned. |
๐งฉ Formation ร Formation Interactions
| 3-4-2-1 vs 4-4-2 | +5 advantage (midfield overload) |
| 4-3-3 vs 3-4-2-1 | +5 advantage (exploits wingback space) |
| 5-3-2 vs 4-3-3 | +2 advantage (defensive solidity) |
| 4-4-2 vs 4-3-3 | +3 advantage (compact counter-attacking) |
| 8 formations ร 8 formations matrix. Cross-multiplied at ร0.08 per advantage point. | |
๐ Bayesian Updating
| Learning rate | 3% per game โ ratings adjust based on surprise (actual โ predicted goal diff) |
| Overperformers | DEF, MID, ATT boosted. Teams winning by more than expected get stronger. |
| Underperformers | Ratings reduced. Engine learns from tournament, not just club seasons. |
โก Momentum Tracker
| Win boost | Winning team gains +0.05 momentum. Losing team loses 0.05. |
| Compound effect | Teams on 2+ game win streaks get escalating advantages. |
๐ Team Scoring (70%)
| Factor | Weight | Source |
|---|---|---|
| DEF โ Defensive strength | Counters opponent ATT | Club stats: tackles, interceptions, blocks, duels won per 90 |
| MID โ Midfield control | ร 0.8 modifier | Club stats: pass accuracy, key passes, dribbles completed per 90 |
| ATT โ Attacking threat | ร 2.5 rescaler | Club stats: goals, assists, shots on target, penalties per 90 |
| GK โ Goalkeeper | Opponent penalty | Overall rating from club season |
| League coefficient | ร 0.85โ1.30 | Normalizes by league quality (PL 1.30, K-League 0.90) |
โก Set Pieces
| Average height | Starting XI average height (taller = better aerial threat) |
| Dead ball specialist | Best MID rating among starters (better delivery = more goals) |
| ร 0.3 weight | ~30% of World Cup goals come from set pieces |
๐ค Cohesion & Chemistry
| League overlap | % of starting XI from the same club league. Higher = better chemistry. |
| ร 1.5 weight | Players who train together weekly outperform individual talent at World Cups. |
๐ฏ Manager Impact (30%)
| Manager rating | Quality, World Cup experience, recent form (scale 5.0โ9.5). Normalized to team scale. |
๐ง Psychological & Environmental
| Factor | Value |
|---|---|
| Tournament pedigree | +0.1 to +0.6 (ARG, FRA, GER, BRA, ESP, ENG, URU, CRO, NED, POR, MEX, BEL) |
| Home advantage | ร 1.15 (host nations), ร 1.03โ1.05 (regional proximity) |
| Altitude | +0.3 (Mexico City 2,200m โ MEX, ECU, COL) |
๐ก Data Sources
| Player stats | API-Football โ club season 2024/25 statistics |
| Lineups & injuries | API-Football โ official matchday team sheets |
| Manager ratings | Manual assessment: quality, WC experience, recent results |
| Live scores | API-Football โ updated every 2 minutes via automated cron job |
๐ Update Frequency
| Live scores | Every 2 min |
| Team ratings | Every 2 min (recalculated as lineups announced) |
| Predictions | Recalculated on page load from latest team_scores.json |
Built in ~1 hour for R49.03 ($3.01) in AI tokens. No subscription. Powered by Hermes Agent (Nous Research) / DeepSeek.