๐ AI Predictor vs Actual โ Accuracy
๐ Today's Games
๐ Predictions lock once the match kicks off. Make sure to enter yours before the scheduled time.
๐ All Group Stage AI Predictions (with actual scores for games already played)
๐ All Group Stage Matches
๐ Group Standings
โญ Team & Player Ratings
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
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) |
๐ค Player Rating Methodology
| Player Quality | Median of all available seasons (going back to 2010). Resistant to outlier seasons โ one injury year doesn't tank a player's true level. |
| Player Form | Most recent club season (2024/25). Captures current momentum. Blended 60/40 with tournament form as WC progresses. |
| Position weights | DEF stats ร3 for defenders, MID ร3 for midfielders, ATT ร3 for forwards. A CB's goals barely count; a striker's tackles barely count. |
| Injury/suspension | Unavailable players excluded. Bench players contribute at half weight. Substitutes included for squad depth. |
๐ Stat Weightings by Position
| Stat | Weight | Rationale |
|---|---|---|
| Defensive Stats | ||
| Interceptions | ร1.5 | Reading the game. Anticipation > reaction. Elite defenders intercept before danger develops. |
| Tackles | ร1.2 | Winning the ball back. Important but can indicate prior positioning error. |
| Blocks | ร1.0 | Last-ditch defending. Valuable but reactive. |
| Duels won | ร0.5 | Physical contests. Noisy stat โ varies by playstyle, not quality. |
| Midfield Stats | ||
| Key passes | ร0.9 | Passes leading to shots. Creativity proxy. |
| Pass accuracy | ร0.6 | Pass completion %. Measures ball retention and reliability. Benefits all positions. |
| Attacking Stats | ||
| Goals | ร3.0 | Ultimate output. Primary job of forwards. Directly changes scorelines. |
| Assists | ร2.0 | Creating goals. Valuable but requires the finish. |
| Shots on target | ร0.8 | Threat proxy. High volume = consistently dangerous even without scoring. |
| Goalkeeper Stats | ||
| Save % | ร9.0 | Core metric: saves / (saves + conceded). A 75% save rate โ 6.75 GK rating. |
| Penalty saves | +0.5 each | Penalty stoppers get bonus. Max +1.0 total boost. Clutched in tournaments. |
| Position multipliers (team contribution) | ||
| D (Defender) | DEF ร3 | MID ร1.5 | ATT ร0.5 | Defending is primary. Passing matters. Goals are a bonus. |
| M (Midfielder) | DEF ร1.5 | MID ร3 | ATT ร1.5 | All-round role. Controls tempo, creates, defends equally. |
| F (Forward) | DEF ร0.5 | MID ร1.5 | ATT ร3 | Scoring is primary. Links play. Pressing counts slightly. |
| Example โ Premier League Forward | ||
| Goals | 11.70ร | 3.0 (goals) ร 3.0 (forward) ร 1.30 (PL) = 11.70 |
| Duels won | 0.33ร | 0.5 (duels) ร 0.5 (forward) ร 1.30 (PL) = 0.33 |
๐ Manager Rating (30%)
Each of the 48 managers is evaluated across four dimensions using career results, league performance, and tournament history.
| Quality Rating | 0โ10 based on career trophies, WC experience, tactical sophistication, and man management. Stored frozen in the app. |
| Form Rating | 0โ10 based on recent results (12 months), win streaks, and pre-tournament momentum. |
| Tactical approach | Documented formation, pressing style, attacking patterns, and defensive philosophy for each manager. |
| Man management | Leadership style, player relationships, handling of star players, and tournament temperament. |
| Examples: | |
| Scaloni (ARG) | Q9.5 F9.0 โ WC + 2ร Copa. Fluid tactics. Exceptional player trust. |
| Deschamps (FRA) | Q9.0 F7.5 โ 3 major finals in 4 tournaments. Pragmatic, elite man-manager. |
| Tuchel (ENG) | Q9.0 F8.0 โ CL winner. Tactical chameleon. Intense, demanding. |
๐ Squad Sentiment & Attitude
| +1.0 Elite | Unbreakable team spirit, players/managers publicly united, "siege mentality" after setbacks. Scaloni/ARG 0.95, Daliฤ/CRO 0.85. |
| +0.5 Good | Positive vibes, training energy high, fans/media supportive. USA 0.60, Norway 0.50. |
| 0.0 Neutral | No notable drama, quiet camp. Ecuador 0.25, Algeria 0.10. |
| โ0.5 Toxic | Manager mutiny, player feuds, media attacks, fan protests. KSA/Mancini โ0.40. |
| Scale: โ1.0 to +1.0. Applied at ร0.30 โ squad mood is decisive. | |
๐ค Recent H2H Form
| Recent win | +0.20 โ Beat this opponent in last 18 months. Confidence edge. |
| Recent draw | 0.00 โ Neutral. No psychological advantage. |
| Recent loss | โ0.10 โ Lost to them recently. Psychological scar. |
| Source: World Cup qualifiers, continental competitions, friendlies โ last 18 months. | |
๐ Squad Depth & Fatigue
| Squad depth | ร0.05 โ 5-subs era rewards deep benches. FRA 9.5, HAI 3.5. Teams with weak benches fade after 60โฒ. |
| Fatigue penalty | (4 โ rest days) ร 0.06 โ Games every 3 days hit performance. 2-day turnaround = โ0.12 penalty. |
| First group game | No fatigue. Fresh teams. Penalty only applies from MD2 onwards. |
๐ Trophy Cabinet โ Champion's Edge
| World Cup | ร1.5 โ Ultimate prize. Argentina's 2022 winners get maximum weight. |
| Champions League | ร1.2 โ Elite club level. Modriฤ 6ร, Bellingham 2ร. |
| Euros / Copa Amรฉrica | ร1.1 โ Continental prestige. Spain Euro 24, France WC 18. |
| PL / La Liga | ร1.0 โ Top 5 league titles. Salah PL, Son Golden Boot (no medal). |
| AFCON / Asian Cup | ร0.9 โ Continental championships outside Europe/South America. |
| Scale: ARG 9.5 ยท FRA 9.0 ยท ESP 8.5 ยท NZL 2.0 ยท HAI 2.0 | |
| Weight | ร0.05 โ Champion's edge. Winners know how to win close games. |
๐ฏ Set Piece Specialists
| Rating โฅ 9.0 | +0.5 โ Elite dead-ball ability. Ward-Prowse, Messi. |
| Rating โฅ 8.0 | +0.3 โ Top-tier specialist. รalhanoฤlu, De Bruyne, Modriฤ. |
| Rating โฅ 7.0 | +0.1 โ Above-average. Son, Salah, Robertson. |
| Stacking: Multiple specialists add together. England: Ward-Prowse (9.5) + TAA (8.5) = +0.8. | |
๐ซ Missing Key Players Penalty
| Star >9.0 missing | โ0.25 (Tier 1 ร1.3, Tier 5 ร0.5). Mbappรฉ out = big impact. |
| Star 8โ9 missing | โ0.15. Tier-adjusted. Less severe for lower-tier teams. |
| Best form >9.0 missing | โ0.20. The in-form player who's carrying the team. |
| GK >8.0 missing | โ0.10. Goalkeeper quality can swing tight games. |
Set starUnavailable:true, bestUnavailable:true, or gkUnavailable:true in data.js | |
๐ฅ Team Intensity
| Pressing actions | 35% โ Tackles + interceptions per 90 across starting XI. Austria (Rangnick gegenpressing) = elite. |
| Physical aggression | 25% โ Fouls committed รท suffered + cards. Uruguay (Bielsa man-marking) = very high. |
| Work rate | 20% โ Average minutes per appearance. Austria avg 90โฒ = relentless. Qatar avg 80โฒ = softer. |
| League tempo | 20% โ Bundesliga/PL players naturally play faster. QSL/Tunisian league = lower tempo. |
| Scale: AUT 9.0 ยท URU 8.5 ยท GER 8.5 ยท QAT 4.5 ยท KSA 4.0 | |
| Weight in Total | ร0.08 โ intensifies pressure differentials. Pressing teams force errors in tight games. |
๐ค Manager-Team Chemistry
| Tenure length | 30% โ Years in role. Scaloni 8yrs (9) vs Dorival <1yr (4). Time builds trust. |
| Win % under manager | 25% โ Results with THIS manager and THIS squad. |
| System fit | 20% โ Does the manager's DNA match the squad? Marsch pressing + Canada counter = poor fit (4). |
| Squad harmony | 15% โ Public conflicts (Mancini/KSA: 2) vs total buy-in (Dalic/Croatia: 10). |
| Continuity | 10% โ Core unchanged? Moriyasu's Japan has played same system for 5+ years. |
| Scale: ARG 9.4 ยท JPN 8.5 ยท CRO 8.4 ยท CAN 4.3 ยท KSA 3.3 | |
| Weight in Total | ร0.08 โ small but decisive in evenly-matched games where chemistry tips the balance. |
๐ฆ Tenacity Factor
| Comeback rate | Games where team trailed but did not lose. Croatia leads at 60% knockout comeback rate. |
| Late goals (75โฒ+) | Percentage of goals scored after 75 minutes. Measures endurance and pressure performance. |
| Tournament overperformance | Actual tournament finish vs FIFA rank / Elo expectation. Morocco 2022: +18 ranks. |
| Penalty shootout record | Historical shootout win %. Croatia, Argentina, Netherlands in top tier. |
| Weight in Total | ร0.10 โ modest but decisive in close games. Croatia (+0.95), Cape Verde (+0.80) benefit most. |
๐ Manager Quality Rating
| Factor | Weight | Description |
|---|---|---|
| Career trophies | 35% | Domestic leagues won, Champions League titles, continental cups, cup competitions. Scaloni: WC + 2ร Copa = 10. Dorival: Copa Lib + Copa do Brasil = 7.5. |
| Tournament experience | 25% | World Cup finishes, continental tournaments managed, knockout progression. Deschamps: 3 major finals in 4 tournaments = 9.5. |
| Tactical sophistication | 25% | System complexity, adaptability, in-game adjustments, pressing/possession philosophy. Bielsa: 3-3-1-3 man-marking = 10. Marsch: Red Bull gegenpress = 5 (poor squad fit). |
| Man management | 15% | Player relationships, squad harmony, crisis handling, dressing-room reputation. Dalic: father figure, players would die for him = 10. Mancini: clashes with federation = 2. |
๐ Manager Form Rating
| Factor | Weight | Description |
|---|---|---|
| Recent win % (12 months) | 50% | Qualifying performance, friendlies, pre-tournament results. Japan: unbeaten since October including England/Brazil wins = 9.5. |
| Win streaks / momentum | 30% | Current unbeaten runs, dominant scorelines, momentum entering tournament. Colombia: 20+ game unbeaten streak = 9.0. |
| Squad stability | 20% | Lineup consistency, injury crises, player availability. Argentina: core unchanged from 2022 win = 9.5. England: Tuchel still experimenting = 6.0. |
๐ How Manager Ratings Feed Predictions
Manager weight varies by team tier โ 25% for Tier 1 (Scaloni's tactics amplified by elite players), 5% for Tier 5 (no manager overcomes massive talent gaps). See Tier-Based Weighting above for full breakdown.
๐ Tier-Based Weighting
| Tier | Team Weight | Manager Weight | Rationale |
|---|---|---|---|
| 1 โ Favourites | 75% | 25% | Scaloni's tactics amplified by Messi's execution |
| 2 โ Contenders | 80% | 20% | Dalic's nous matters but players decide |
| 3 โ Dark Horses | 85% | 15% | Manager impact limited by squad ceiling |
| 4 โ Outsiders | 90% | 10% | Tactics can't overcome talent gaps |
| 5 โ Long Shots | 95% | 5% | No manager overcomes massive differentials |
โฐ Peak Too Early Regression
| Knockout rounds only | Teams that overperform in groups get a regression penalty. Backed by historical data. |
| Overperformance = actual GD โ predicted GD | +6+ overperformance โ โ0.40 penalty. +4 to +6 โ โ0.25. |
| Late bloomers boosted | Underperforming in groups โ +0.08 to +0.15 boost in knockouts. |
| History | 6 of 7 group dominators since 2002 lost before SF. Spain '06 (+7 GD โ R16 exit), Argentina '10 (+6 GD โ QF 4-0 loss). |
๐ Home Advantage & Host Boost
| Host nations | USA ร1.15, MEX ร1.15, CAN ร1.10 โ crowd, travel, and familiarity advantages. |
| Regional proximity | CONCACAF teams playing in North America get ร1.03โ1.05 bonus. |
| Altitude venues | Mexico City and Guadalajara give MEX, ECU, COL +0.30 altitude edge. |
| Host tenacity | Host nations get ร1.2โ1.3 multiplier on tenacity score. Crowd energy in pressure moments. |
๐ Full Formula Reference
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.