Methodology

Each team has a score that reflects its relative strength. After every match, scores are adjusted based on the result, the scoreline, and the difference in strength between the teams. Home teams receive a bonus reflecting their home advantage. At the start of each season, scores are partially regressed toward the mean to prevent past seasons from weighing too heavily.

To predict the outcome of each future match, we use the Elo difference between teams to estimate how many goals each should score. From these estimates, we calculate the probability of every possible scoreline. We apply a statistical correction that improves accuracy for low-scoring matches, where the base model tends to be less precise.

We simulate the rest of the season thousands of times. In each simulation, we draw a scoreline for every future match based on the calculated probabilities, update the standings, and record each team's final position. By the law of large numbers, the more simulations we run, the closer the results converge to the true probabilities. In practice, after a few thousand iterations the probabilities stabilize with high precision.

For each objective (title, qualification for international competitions, relegation), we calculate how many wins a team needs to achieve it at different confidence levels. We force the team to win a specific number of matches and check in how many simulations it reaches the objective. The minimum number of wins that guarantees the objective in 75%, 90%, or 99% of simulations is what we call Wins Needed.

Data is updated automatically after each round. The pipeline fetches results, recalculates ratings, runs simulations, and publishes the new numbers on the site.