Digia released an AI-based forecast for the winner of the ice hockey SM-league last November. The forecast was updated at the end of the year, with Tappara, Pelicans, TPS, and Ilves rising to the top four. In the latest update, the top duo has remained the same, while HIFK and Kärpät have risen to fight in the bronze match.
When about two-thirds of the regular season is behind, it seems clear that certain things, such as the success of the Pelicans, the AI has seen more accurately throughout the season than human predictors. On the other hand, changes in team lineups and other surprises have shaken the settings in a way that was not readable from the data at the beginning of the season.
The differences in human and machine views are exacerbated in Ilves predictions
A fresh example of surprise changes is defender Sami Niku’s move from JYP to Ilves. Ice hockey expert Petteri Sihvonen predicted Ilves to be the champion already last year, and still sticks to his prediction. Sihvonen tells in a video interview, how in addition to other positive signals, he believes Niku’s transition will bring Ilves a boost that will propel the team to the championship - even though the cat’s “fur has been tangled” recently.
The AI’s assessment of the prospects of the Tampere team is harsher. In the original forecast, the AI raised Ilves to be the champion, but at the moment Ilves would end up in fifth place.
The forecast is based on data from previous games. The original forecast published last year was almost entirely based on last season’s (2022-23) data. As the season has progressed, the model has been supplemented with fresh data. Therefore, the drop in Ilves’ ranking in the forecast is simply based on this season’s game success, which according to the AI’s calculations does not give reason to expect medal positions.
Niku’s transition is such fresh news that it does not yet appear in the current data. What then is the impact of an individual player on the team’s overall performance?
“In human thinking, the impact of individual events or even individual hero players is often greater than it actually is. The AI model takes into account dozens of variables, which balances the impact of individual factors on the forecast,” says Juhana Juppo, CTO of Digia.
“On the other hand, one of the strengths of AI is precisely the identification of influential factors from huge data masses. For example, when a company manager - or a hockey coach - needs to identify what things to develop to get the most impact, AI provides very good tools for such impact analysis,” Juppo says.
Petteri Sihvonen’s emotional world is in turmoil - AI remains cool
With the SM-league forecast, Digia wants to illustrate the possibilities of utilizing AI.
“There is now a lot of hype and misconceptions associated with AI, even though in reality it is an everyday tool like other technologies. It is just necessary to understand where it is most beneficial, and where some other tool helps more. Understanding the strengths and limitations of AI is the key to its effective use,” says Juhana Juppo.
The best outcome in ice hockey as well as in other areas of life is achieved by combining the computing capacity of AI and human intuition. AI has almost unlimited computing power on its side, but humans can also take into account things outside the data.
“Recently, a lot has happened in the regular season that shakes my emotional world - in that sense I am not as cool as AI,” Petteri Sihvonen ponders in a video interview.