The hockey simulation — the supercomputer plays the season a million times
Who wins the SHL? Who goes up from HockeyAllsvenskan? Instead of predicting a table, we let the supercomputer play the whole season a million times per league and count how often each team finishes in each position. Here are the ways into both leagues.
SHL simulated 1,000,000 times
- 1
Skellefteå AIK37.7% - 2
Frölunda HC26.5% - 3
Brynäs IF16.9%
HockeyAllsvenskan simulated 1,000,000 times
- 1
Kalmar HC41.3% - 2
BIK Karlskoga23.1% - 3
Leksands IF18.4%
How the supercomputer works it out
Every match is replayed
Each remaining match is decided at random from the teams’ estimated strength — home advantage, goalscoring and defending weighed against the opposition. No match is a foregone conclusion, just as in real life.
A million seasons
The whole league is played through a million times. Points are awarded under hockey’s system: three points for a win in regulation time, two and one when the match is decided in overtime or a shootout.
The distribution is the answer
Once every season has been played, we count how often each team won the league, made the play-offs or ended up in the qualifier. That is the team’s probability — not a guess, but a frequency.
The model’s baseline forecast is supplemented with editorial strength adjustments ahead of the season (squad changes).
Why simulate instead of predicting a table?
A predicted table before the season is a single story about how things might go. A simulation is a million stories. The difference is clearest in mid-table, where four or five teams are often practically equal: a predicted table has to rank them, while the simulation shows that they have roughly the same chance — and how big it is.
That is also why the simulation is useful for betting. When a site writes that a team "should" finish third, it says nothing about the odds. When the simulation says the team wins the league in 12 per cent of seasons, you can compare that directly with the bookmaker’s price on the league title — and see whether there is value or not.
The same method as for football
We have simulated football leagues this way for several seasons — from the Allsvenskan to the Premier League. Hockey needs its own adjustments: the points system with overtime, the larger home advantage and the more volatile scorelines. But the principle is exactly the same, and the predictions before every match come from the same model.
Frequently asked questions about the hockey simulation
How does the SHL season simulation work?
The supercomputer plays the whole remaining season a million times. In each run every match is decided at random based on the teams’ strength, and the points are recorded — including hockey’s points system with three points for a win in regulation time and a shared pot when a match is settled in overtime. Once a million seasons have been played, the distribution shows how often each team won the league, made the play-offs or ended up in the qualifier.
What does the probability of winning the league mean?
A team with an 18 per cent chance of winning the league won it in roughly 180,000 of a million simulated seasons. It is not a forecast that it will happen — it is a measure of how big the chance is given what we know today.
Do you simulate HockeyAllsvenskan too?
Yes. The same model runs on HockeyAllsvenskan, where the race for direct play-off places and the promotion qualifier makes the simulation at least as interesting as the title race in the SHL.
How often is the simulation recalculated?
After every round played. New results change both the table and the teams’ estimated strength, so the probabilities move throughout the season — most of all where a few points decide.