90 likes | 113 Views
Develop an algorithm to forecast NBA player statistics for upcoming games by considering factors like pace and defensive metrics to enhance accuracy. The algorithm processes defensive and offensive team data along with individual player performance indicators. Challenges include real-time calculations, varying statistics, and accuracy hurdles similar to predicting NCAA games. Employ Brute Force and Dynamic/Constrained algorithm types for analysis, yielding results within one standard deviation. Success in this task is gauged by the precision of predictions. Restricting the problem scope can enhance accuracy or efficiency, like in the Traveling Salesman Problem.
E N D
NBA Analytics Zyrus Johnson
Problem Inspiration • https://youtu.be/Mq785nJ0FXQ?t=11m26s
Problem Description • Create an algorithm that predicts NBA player stats for an upcoming game • Factor things such as pace and defensive statistics in order to get as good of an estimate as possible
Formal description • defStats(defTeam, offTeam, offPlayer) • Defensive and offensive teams are lists including defensive rating, pace, steals, and defensive rebounding percentage • Offensive player is a list of points, rebounds, assists, steals, and turnovers
Challenges • This is a calculation that does not happen in real time • Statistics vary and there is no way to calculate things such as motivation • Accuracy is difficult like predicting NCAA tournament games
Algorithm types involved • Brute Force - For each new player there were two lists from different websites to synthesize. This was done by going alphabetically through players and analyzing them • Dynamic/Constrained-This problem was constrained by using the first list (all the players in the NBA) and returning only those who played 20 or more minutes a game. Dynamically limiting the sample made the problem easier and more accurate
Results • This is like a limit problem since it’s impossible to get to the exact correct answer • This program gives the statistics within one standard deviation
Question 1 • Simply because a problem isn’t complex, doesn’t mean it’s accurate. In a problem such as this, what specific ways would success be measured? • This type of problem would be successful based on accuracy of predicitons.
Question 2 • A program can either be judged based on accuracy (predictions like the weather) or efficiency (traveling salesman). Give an example of when constraining the problem would make it more accurate or more efficient. • In the TSP problem shortest path gives you less operations than O(N!) which is more efficient.