About the model
How Our World Cup 2026 Prediction Model Works
Every prediction on this site is produced by the same six-factor weighted model. Here's exactly how it works, what data it uses, and why we weight things the way we do.
Model Summary
The model combines team strength (35%), player quality (25%), coaching and tactics (15%), momentum (10%), path difficulty (10%) and chaos control (5%) to produce win-probability estimates for every match and tournament outcome.
The Six Factors
| Factor | Weight | What It Measures |
|---|---|---|
| Team Strength | 35% | Collective squad quality, tactical organisation and defensive/attacking balance based on recent performances and Elo ratings |
| Player Quality | 25% | The individual ability of key players — starters and depth — weighted by how likely they are to play and at what level |
| Coaching & Tactics | 15% | Managerial tenure, tactical coherence, in-game management quality and historical tournament performance under each coach |
| Momentum | 10% | Recent form (last 10 competitive matches), qualification campaign quality, and team morale indicators |
| Path Difficulty | 10% | The specific opponents a team is projected to face — easier paths increase expected tournament depth |
| Chaos Control | 5% | Tournament-specific unpredictability: injury risk, suspension accumulation, penalty shootout probability |
Data Sources
The model is built on publicly available data combined with analyst judgment. The primary sources are:
- FIFA World Rankings — the official ranking gives a baseline for each team's global standing.
- Club Elo ratings (club-elo.com) — aggregated by squad to give a more granular player-quality picture than international rankings alone.
- Recent international results — last 10-12 competitive internationals weighted by opposition quality.
- Squad depth analysis — quality of first 11 vs. 23-man squad, injury updates and suspension status.
- Managerial tenure data — longer-tenured coaches with stable systems typically outperform in tournaments.
- Historical tournament data — head-to-head records in World Cup knockout stages and major tournaments.
Why We Weight Team Strength Highest
Team strength — the collective — accounts for 35% because football is a team sport. Individual brilliance (Mbappe, Haaland, Ronaldo) can change a single game, but it doesn't reliably predict tournament outcomes as strongly as the team's collective defensive organisation, pressing intensity and set-piece quality.
This is why we have Spain at the top despite Argentina having Lionel Messi as a factor. Spain's team strength score — defensive structure, pressing system and midfield control — is marginally higher than Argentina's individual-centric model, and over six or seven games, that collective edge accumulates.
How Match Predictions Are Made
For each individual match prediction, we run both teams through the model to get a team score, then calculate win probability using adjusted Elo difference with a home/neutral advantage correction. For World Cup knockout stages, we apply an additional weight for tournament experience and penalty shootout resilience.
The resulting probability distribution is:
- If win probability >65%: we predict a comfortable win (e.g., 2-0, 3-0)
- If win probability 55-65%: we predict a narrow win (e.g., 2-1, 1-0)
- If win probability 45-55%: we predict a draw or very narrow win for the slight favourite
Limitations and Caveats
Every model has blind spots. Ours are:
- Injuries in-tournament: a key player injury during the tournament dramatically changes probabilities but can't be predicted.
- Tactical surprise: a manager deploying an entirely new system in the tournament that the model hasn't seen in the data.
- Psychological momentum: the model captures recent form but can't fully quantify the psychological impact of a dramatic comeback or devastating loss.
- Penalty shootouts: tournament results often hinge on penalties, which are essentially unpredictable beyond a narrow band of historical shootout performance.
All predictions on this site are analytical opinions, not certainties. We are modelling probability distributions, not stating outcomes. Use them as informed starting points for discussion, not guarantees.