Abstract
The various events occurring during a basketball game such as shots made and shots missed, and the factors affecting these events such as the time period when they occurred or the players on the court when the event takes place, can build up more or less regular patterns that can eventually be defined by player, team and/or year. These patterns can exist in terms of coördinate values on a typical basketball court, be it on a Cartesian coördinate system or that of a Polar coördinate system. These can be used to employ algorithms that can be helpful during machine learning to predict the events of each game by player and by team. Subsequently, the outcome of these events can be determined and evidences to known and unknown truths about basketball can be provided. The learning process requires the collection of previous game events by player, team and year; the organization and representation of this information in the most appealing ways to determine the existing patterns (data mining), and consequently the algorithms to be used for the machine learning experiences. However, the data mining method used remains one of the most important decisions for the success of this method. Data analysis by visualization and the analysis of data variations as continuous events rather than discrete ones are especially useful for this process, as it simultaneously interprets various data and provides evidences to this interpretation.

Keywords: data mining, machine learning, basketball, games