Football Sim Lab
Simulation methodology and data
Last reviewed:
This describes the tournament engine shipped with the site. Player Career uses separate fictional progression and selection rules.
Ratings and data provenance
The 2026 preset has 48 teams with bundled Elo-style numeric ratings. These are model inputs, not a live FIFA feed. The retained table lacks a verifiable source date for each rating, so we cannot certify it as a current or exact historical World Football Elo snapshot. FIFA ranking positions and Elo rating points are different quantities.
The 2030 preset uses a bundled ranking order originally labelled 20 July 2026, regional quotas and configured hosts. The original import recorded a secondary transcription and a FIFA URL with a different date parameter. Its ranking provenance has not been independently verified. Treat it as a game projection, not confirmed qualifiers or a verified forecast.
Consult FIFA's ranking publications for official positions and World Football Elo Ratings for general Elo background. These links do not certify our input table.
Win, draw and loss
The starting comparison is E(A) = 1 / (1 + 10^((rating B − rating A) / (400 × chaos scale))). An enabled host advantage adds 50 rating points before a match. Standard Chaos Factor 1 maps to a win-probability scale of 0.65. This is game tuning, not a measured accuracy claim.
A separate draw probability starts at 28% for equal ratings and decreases with rating difference, with a baseline floor of 12%. Low chaos further reduces draws. The remaining probability is split between the teams using the Elo comparison.
At standard chaos, two non-host teams rated 1800 have 36.0% first-team win probability, 28.0% draw probability and 36.0% second-team win probability. The example is calculated directly by the shipped engine.
How scores are generated
The engine chooses an outcome first. Draws use weighted 0–0, 1–1 and 2–2 scores. Wins use conditional Poisson sampling with a 2.6-goal baseline, up to 20 retries and a fallback score that preserves the winner. This is not a pure independent-goals Poisson forecast. Knockout ties use the game's extra-time and penalty rules.
Rules and fixed results
Group points, sorting, third-place qualification and knockout assignments follow the implemented rules. Manual scores and optional stored real results can replace generated scores. A fixed result is not a successful prediction. The real-results dataset is bundled and may be incomplete; it is not a live feed.
Official format information is available through FIFA's tournament resources. Report discrepancies through Contact.
Repeated simulations and uncertainty
One run gives one possible story. A repeated-run percentage is a frequency under specific inputs, locked results and model version. A reproducible report should retain those, the run count and random seed. With 1,000 independent runs, the maximum approximate 95% sampling margin is ±3.1 percentage points. That measures sampling noise, not model accuracy.
Archived articles without retained inputs are not verified probability estimates. We have no published backtest demonstrating calibration or superiority to other models. Injuries, lineups, weather, tactics, transfers and betting prices are not continuously incorporated. Use these games for exploration, not financial or betting advice.
Try the worked example or group-stage simulator.