preview

What Are Basketball Statistics?

Decent Essays

Term Project – Basketball Statistics Ever since its inception, basketball has been a difficult sport to quantify compared to others, other than baskets and assists. This is because, unlike a sport such as baseball or football, there are no individual sequences (such as a pitcher throwing a strike, or a running back scoring a touchdown). Instead, basketball revolves around constant flow and ball movement, with play never stopping unless the ball falls out of bounds or a foul is committed. In addition, the rapid movement of the ball ricocheting off players and the basket has made movement difficult to track. One of the most difficult to track has been the probability of a rebound after a missed shot. A rebound occurs when a player shoots, misses, …show more content…

After a missed shot attempt, players from both teams are free to grab the basketball to gain possession. During the 2013-2014 National Basketball Association (NBA) Regular Season, players missed well over 100,000 shot attempts, and since you cannot score without possession, rebounds are extremely valuable in the sport. Teams in the NBA have traditionally relied on natural instincts, and a bit of luck, when battling for rebounds. However, recent improvements in camera technology have started a new analytical age. Because of advanced tracking cameras and the data sets they provide, previously unquantifiable statistics such as which players are the most agile or which players pass the most are available. These cameras, called the SportVU System, are also tracking the precise position of the basketball at all times. This means that not only can the exact location of each missed shot be logged, but the corresponding location of each rebound. Thus, the age-old question of “where are rebounds the most likely …show more content…

While the predictability starts out near 77% when the ball leaves the offensive shooter’s hands, the probability that we can predict the correct team who will make the rebound increases to 87.5% by the time the ball has fallen back down to the 8-foot level, which is the altitude at which most players can begin to grab and control the ball. The support vector machine (SVM) is a state-of-the art machine learning technique that has been shown to work well in a wide variety of settings, with the important trait of avoiding over-fitting. For this data, we used [ALEX, YOU KEEP SAYING ‘WE’. WAS THIS A GROUP PROJECT?] an implementation of an SVM provided by the WEKA machine-learning program. The SVM is able to outperform simple logistic regression by several percentage points in terms of accuracy of its predictions. The graph below shows performance of an SVM with radial basis function (RBF) kernels, which enables the SVM to separate between offensive and defensive rebounds using a hyper-plane in a higher-dimensional space created using this non-linear kernel. Hence the SVM can create non-linear class separators when the solution is projected back down to the original search

Get Access