Every title race reaches the point where someone posts a percentage. A club is "63% to win the league", another has "a 9% chance of going down". Those numbers look precise, and people argue about them as if they were predictions. They are something more modest and more interesting: the result of playing the rest of the season thousands of times on a computer.
Here is how that works, step by step, and where it breaks.
Step One: Rate Every Team
A simulation is only as good as its idea of how strong each team is. The simplest rating is the league table itself, but points are noisy. A side that won three games with late deflected goals looks better on the table than it really is.
Most serious models build a rating from a mix of signals:
- Results, usually with recent matches weighted more heavily than ones from months ago.
- Goal difference, which carries more information than points because it separates a 1-0 win from a 4-0 win.
- Expected goals (xG), which measures the quality of chances created and conceded rather than how many went in.
- Home advantage, which in most leagues is worth a fraction of a goal per match.
xG matters because finishing is streaky. A striker who misses three clear chances in a row is usually not a worse striker than last month. A model that leans partly on xG does not overreact to a lucky or unlucky run, and that is why a club sitting fifth can still be rated as the strongest side in the league.
Step Two: Turn Ratings Into Match Probabilities
Once each team has an attacking and defensive rating, the model estimates how many goals each side is likely to score in a specific fixture. A common approach treats goals as a Poisson process: if a team is expected to score 1.6 goals, the model can work out the probability of it scoring 0, 1, 2, 3 or more.
Combine the two teams and you get a grid of possible scores, and from that grid three numbers:
| Outcome | Example probability |
|---|---|
| Home win | 52% |
| Draw | 25% |
| Away win | 23% |
That is a single match. A title race has dozens of them still to play, and they interact. A rival dropping points next weekend changes what every later result means.
Step Three: Play the Season Thousands of Times
This is the Monte Carlo part. The model takes every remaining fixture and, for each one, rolls a weighted die using the probabilities above. A random number decides whether this particular simulated match ends 2-1, 0-0 or 1-3. It then adds the points to the table, moves to the next fixture and keeps going to the final day.
At the end it records one complete final table. Then it starts again from today and does it all over, with fresh random numbers.
After 10,000 simulated seasons you count:
- How many times each club finished first. That share is its **title probability**.
- How many times it finished in the European places. That is its **Europe probability**.
- How many times it finished in the relegation zone. That is its **relegation probability**.
If a club won the league in 6,300 of 10,000 runs, it is "63% to win the title". Nothing more mysterious than that.
Why 10,000 Runs and Not 100
The number of runs controls how stable the percentages are, not how accurate the model is. With 100 runs a club near 50% could come out anywhere from roughly 40% to 60% purely by chance. With 10,000 runs the random wobble shrinks to about one percentage point. Beyond that, extra runs add precision the model has not earned, because the team ratings themselves are far less exact than one percent.
What "What If?" Scenarios Show
The most useful feature of a simulator is not the headline number. It is the ability to fix a result and see what changes. Lock in a derby as a home win, rerun the remaining season, and the title odds of three clubs move at once.
This answers questions fans actually argue about:
- How much does next week's top-of-the-table clash really matter?
- If the leaders drop points twice, is the chasing side favourite?
- What is the minimum a struggling club needs from its last five games to stay up?
Often the answer is surprising. A match that feels decisive moves the odds by a few points, while a dull-looking away game against a mid-table side turns out to matter more because of who else is playing that weekend.
Where Title Odds Models Go Wrong
Simulations are honest about randomness but blind to a lot of football. The common failure points:
- Injuries and suspensions. A model rating a team from season-long data does not know the first-choice striker went off in training yesterday.
- Transfers. A January signing or departure changes a squad faster than results can show it.
- Motivation and fixture congestion. A club with nothing to play for in May, or one rotating for a cup final, is weaker than its rating.
- Managerial changes. A new coach can shift how a side plays within two or three matches.
- Small samples early in the season. After six rounds, ratings lean heavily on last season and on assumptions.
So a 63% title chance means "if the ratings are right and nothing unusual happens". That is a useful frame for a debate. It is not a forecast you should stake money on, and bookmaker prices, which carry their own margin and far more information, are a different thing entirely.
Run It Yourself With Simulig
We built Simulig to put this whole process in your pocket for the Turkish Super Lig. It starts from the current table and plays every remaining fixture 10,000 times, with team strength calculated from real results and xG. You see each club's title, Europe and relegation probability, and "What if?" lets you set the score of any remaining match and watch the table recalculate instantly. The table and squads are refreshed every week.
It goes further than the odds table. There is a manager career mode with the real squads, a live 2D match where all twenty-two players move, quick matches between 170 clubs, 52 national teams and 50 legendary squads, and World Cup and European Championship tournament formats.
Two honest limits. First, Simulig shares the blind spots described above: it does not know about yesterday's training ground injury, so treat its odds as a well-founded starting point for an argument, not a prediction. It is for entertainment and is not betting advice. Second, the app is in Turkish and is currently available only on the Turkish App Store. If you follow the Super Lig and can read Turkish, it is free to download, with an optional Simulig Pro subscription for the full career, scenarios and tournaments.
The Short Version
A title percentage is the share of thousands of simulated seasons in which a club finishes first. Each simulated match is drawn from probabilities built on team ratings, and good ratings blend results with xG. The method is sound and the arithmetic is simple. The weak point is always the ratings, because football changes faster than any model can see. Use the numbers to understand a title race, and enjoy the argument.