Artificial Intelligence Surprises in Digia’s Liiga Forecast – How Did the Prediction Fare in the Regular Season?

The race for the Finnish ice hockey championship is heating up. With the regular season now sealed, only ten playoff teams remain in contention. AI and hockey analyst Petteri Sihvonen hold differing views on how the games will unfold.

Sihvonen believes that Ilves will clinch the championship. In his forecast, HIFK secures the second spot, while Tappara ranks third. However, the AI forecast places Tappara, Pelicans, and Lukko at the top.

“This actually challenges my entire role as a hockey analyst because I’ve publicly expressed my lack of trust in Tappara. Now we’ll see who’s right regarding Tappara – me or the AI,” says Sihvonen.

“This has made it incredibly exciting. My blood is pumping,” adds Sihvonen.

The divergence in the AI forecast can be attributed to different outcomes in the tightly contested quarter-final series compared to Sihvonen’s predictions.

You can explore the playoff bracket on Digia.com, where each team’s projected outcome is based on game-by-game results. The forecast will be updated as the playoffs progress.
 

How did AI fare in predicting the regular season?

Now that the regular season is wrapped up, we can assess how well AI performed in its predictions. In many instances, the AI was quite accurate, although it didn’t foresee everything.

•    The model correctly predicted the regular season winner and the second-place team.

•  It estimated a two-point gap between the first and second positions, whereas the final difference was three points.

•    Artificial Intelligence Predicted Nine out of Ten Playoff Teams Correctly: While many comparative forecasts hit the mark for eight teams, AI accurately predicted nine out of ten playoff teams.

•    Jukurit’s Fifth Place Was Less Surprising for AI: Contrary to others, AI originally forecasted Jukurit to finish in the 7th position, whereas other predictions had Jukurit ranked 12th or 13th.

•    Kärpät Outperformed the Prediction: The forecasted placement for Kärpät was 10th, but they impressively secured the fourth position.

According to Digia’s Chief Technology Officer, Juhana Juppo, the model performed well overall in predicting the regular season.

“The model is based on historical game data. However, surprises can always occur that are not evident from historical data. For instance, Kärpät changed coaches mid-season, resulting in a completely different playing style,” says Juppo.

This type of AI forecast yields good results when there are numerous variables, and the goal is to determine which factors are most crucial for the final outcome.

“In the Liiga forecast, we modeled hundreds of different factors that could potentially impact the winning team’s prediction. From these factors, we selected around 20 variables that had the most significant influence on the outcome. Now that we can compare the forecast to the regular season results, we consider the modeling successful,” Juppo explains.

Analyzing the model’s successes helps us understand AI’s strengths and weaknesses more broadly.

“AI is a tool for everyday use, just like other technologies. The key lies in understanding where it provides the most benefit and where other tools may be more helpful. Appreciating AI’s strengths and limitations is essential for effective utilization,” concludes Juhana Juppo.


More information:
Juhana Juppo
Chief Technology Officer, Digia
tel. 040 172 2859
juhana.juppo(a)digia.com