World Cup 2026 Group Winner Modelling: Retrospective Method

Status of this page

This is a retrospective methodology page, not a live 2026 group tip sheet. Use FIFA's official schedule and results for completed outcomes and its format and tie-break guidance for the competition contract.

Pre-tournament simulation contract

For every simulated group, calculate points and apply the official tie-break criteria in order. Do not rank tied teams alphabetically or by model strength unless the rules explicitly reach that state.

Distinguish event probabilities

P(team wins group) is the sum of complete tournament paths in which the team finishes first. It is not the average of match win probabilities, the chance of qualifying, or the probability of winning the opening match.

Group matches are dependent through table state and incentives. OpenStax's independence condition prevents treating all tournament paths as products of unchanged marginals when later match probabilities depend on earlier outcomes.

Retrospective evaluation

Recreate the original forecast from its saved model and data cutoff. Do not insert final squads, lineups or results. Use chronological evaluation under scikit-learn's TimeSeriesSplit guidance for model development and calibration guidance for probability reliability across comparable group forecasts.

Compare forecast probability with the accepted historical price only if that price and contract were archived. A correct winner does not by itself validate the probability.

Next step

Use World Cup 2026 Accumulator for the next part of this topic.

Reproducible group simulation

Using FIFA's official group and tie-break rules, verify these invariants before publishing any retrospective or future group probability:

• Every scheduled group match appears exactly once.
• Win and draw points reconcile at team and group level.
• The official tie-break sequence is applied in order.
• Team finishing-position probabilities each sum to one within declared tolerance.
• Mutually exclusive group-winner probabilities sum to one.
• The random seed, number of runs, input timestamp and code version are recoverable.

Under time-ordered validation guidance, backtest only with inputs that existed before each simulated match. Final scores, later squad news and completed group standings cannot be features in a claimed pre-match forecast. Compare the model with simple declared baselines and publish uncertainty, not just the most likely team.

FIFA's official schedule and results show that, as of 15 July 2026, the group stage is no longer an upcoming prediction market. The useful durable output is therefore the method, archived inputs and evaluation record, not a rewritten list of supposed pre-tournament tips.

Continue learning

• Next guide: Boxing Day Accumulators
• Related guide: Opening Weekend Accumulators

Assumptions and limitations

This page reconstructs a method and does not rank 2026 teams. Tournament formats, tie-breaks and participants can change before future World Cups, so future use requires new official sources.