Student of the Department of System Programming of the School of Electronic Engineering and Computer Science Mikhail Medvedev has created the first in Russia machine learning model for evaluating the skill level of hockey players. In his work, Mikhail used statistical data not only of the Kontinental Hockey League, but also of other Russian (All-Russian Hockey League, Junior Hockey League) and world leagues (National Hockey League, American Hockey League).
In order to evaluate a hockey player’s efficiency, dozens of indicators are considered, such as: playing time on the ice, accuracy and number of passes, frequency and conversion of shots, hits and checks, penalties, won one-on-one battles, and more. Mikhail used statistics for three seasons and divided players into three categories: defensemen, wingers, and centre forwards. When marking, in each of these categories a hockey player got evaluated as an attack or defense player.
“Technically, there are six machine learning models: two for defensemen (attack and defense) and two models each for wingers and centre forwards, which allow us to comprehensively assess the player's level,” says Mikhail Medvedev. “The Pandas and Sсlearn libraries were used to work with the data. When marking the data, each player received one of the following ratings: A (dominant player), B (high-class player), C (average-level player) and D (low-level player). The created data sets by positions were mixed, then each data set was divided into training, testing and validation. After that, the machine learning models were trained and tested to classify players.”
At all stages of the work, the Chelyabinsk student was assisted by staff members of the Spartak Moscow Hockey Club, where Mikhail has been working for several seasons. The management of the Moscow club noticed the talented programmer and invited him to work as an analyst.
“The system for assessing the effectiveness of hockey players will allow us to objectively assess players," says Mikhail Medvedev. “With the help of artificial intelligence, it will be possible to select athletes with the necessary qualities and strengths for a team. The system also shows how much a hockey player influences the team’s play: how many passes he makes, how often and accurately he shoots, and so on.”
This system has been successfully implemented in the work of the sports department of HC Spartak and is used to compare and search for players not only within the Kontinental Hockey League, but also in other Russian and world leagues. Now the developer is planning to improve the software package and implement other indicators into it that will make the description of players more accurate. And one of the ideas for the future is to teach artificial intelligence to evaluate not only the current level of a player, but also his potential.
“Many believe that artificial intelligence will replace humans in the future. However, our goal is not to replace humans, but to provide maximum information in an accessible form so that in the future the most objective and correct decision can be made by humans. Artificial intelligence should help us and not compete with us,” says Mikhail Medvedev.