It’s a great day in sports analytics! Today we’ll discuss regression and classification and how they are applied in basketball. Game-related statistics that discriminate between winning and losing teams have received plenty of research attention within performance analysis in basketball. Regression and classification has identified consistent key indicators (i.e.. successful 2-point shots, defensive rebounds and assists) that discriminate between teams’ performance in men’s and women’s basketball. Therefore, situational variables have to be taken into account when discriminating between winning and losing teams based on game-related statistics in team sports.
A classification tree analysis is used to classify winning and losing teams according to situational variables and game-related statistics of the slower- and faster-paced games. There are classification trees and decision trees. Decision trees are often used to visualize the different choices, the uncertainty of the choices, and their outcomes. They are easy to visualize and for all audiences. It should be noted that classification trees are not the most efficient algorithm for learning. State-of-the art algorithms, such as random forest, gradient boosting, neural networks, etc tends to perform significantly better. However, classification trees have a huge advantage in terms of visualization and ease of understanding.
Here is a small comparison between Classification and Regression:
Classification is the task of predicting a discrete class label algorithm as a continuous value. But the continuous value is in the form of a probability for a class label. Classification predictions can be evaluated using accuracy, whereas regression predictions cannot. Regression is the task of predicting a continuous quantity, the algorithm may predict a discrete value, but the discrete value in the form of an integer quantity. Regression predictions can be evaluated using root mean squared error, whereas classification predictions cannot.
The use of Regression and Classification in basketball games:
Basketball is the second most popular team sport worldwide, and the second most watched Olympic sport, with over 450 million registered players. The key physical and physiological characteristics of basketball athletes have been documented and reported to contribute to individual performance with team success reliant on the coherent integration of individual performances. Many studies have examined the importance of team performance indicators for
wins within national junior and senior competitions. Most have identified field-goal percentage, defensive rebounds, and assists as crucial team indicators for winning. Recently, these results were extended to the elite international level with field goal percentage and defensive rebounds identified as vital for match outcomes at the men’s basketball.
It is really important to check the difference between slower and faster paced games. In faster paced games for winning teams the tree model showed the predictive importance of assists, successful free-throws, successful 2-point field-goals, fouls committed, and defensive rebounds. These Variables were related to fast-paced rhythms that allowed for better offensive (i.e., quick attacks and high shooting accuracy) and defensive actions (i.e., defensive rebounds). This playing style generates more opportunities for fast breaks when securing defensive rebounds, passing and assisting to an open player in easy field-goal positions without defensive pressure. However, the significant effects of successful free-throws and fouls committed are inconsistent within available research. On the one hand, faster-paced games involve better performance of offensive rebounding due to the fact that this variable secures a restarted ball possession. On the other hand, defensive readiness may be expected during fast-paced games {e.g.. steals and recovered balls instead of fouls committed). While in slower-paced games the tree model showed the importance of assists, successful free-throws, successful 2-point field-goals, fouls committed, and defensive rebounds. In addition, successful 3-point field-goals were included as a significant predictor when classifying winning and losing teams. These key game-related statistics are associated with a high degree of the ball control playing style that is characterized by making better field-goal selection decisions, accurate passing to teammates in open positions, and drawing fouls. Slower-game pace is connected with shooting efficiency, particularly from the mid- and long-range distances (e.g., 2- point and 3-point field-goals). This playing style enhances the importance of selecting good positions for field-goals. In this respect, coaches are well-advised to prepare their games and competitions accordingly.
Both approaches are very useful, but on a case-by-case basis, choosing an algorithm is a critical step in the machine learning process, so it’s important that it truly fits the use case of the problem at hand.
—————–
How mature is your team’s analytics program? Take the Sports Analytics Maturity Assessment.
Learn about the Groundbreaking Sports Analytics Model coaches and sports analysts are talking about!
Learn all about Sports Analytics here.
