March Madness Tournament Case Study

For this project a metric would be used to predict who was going to win the 2017 March Madness Tournament. The metric consists of a team’s RPI and two stats that have a strong correlation coefficient to how the teams finished in the 2016 tournament. The Rating Percentage Index (RPI)is a ranking system that is used to determine a team 's rank in the tournament and during the regular season. The two stats had to have a moderate correlation to how the last 16 teams in the 2016 tournament finished. Also a scatter plot was used to display two different stats and how they correlated. Once the two stats were determined each part of the metric(RPI and two stats) have to be weighted by percent. For example, a metric can be RPI times 0.3 + stat1 times 0.3 + stat2 time 0.4. Then the metric is to be used with every team in the 2017 tournament. Now on the 2017 bracket the team with the higher number for the metric should win the game. Lastly do this for every game and a winner will be decided. After trying over 10 stats to find two that have a moderate correlation, only two of them were close. Finding a strong correlation is very difficult. This is because of all the little factors that

Then the championship is SMU vs. Purdue and then SMU is supposed to win it all. The metric that was used was RPI times 0.2 + Opponent’s miss percentage times 0.41 + Assists per field goal made time 0.39. The reason the two stats were opponent’s miss percentage and assists per field goal made because those were the closest stats to a moderate correlation. The opponent’s miss percentage had a correlation coefficient of 0.48 and the assists per field goal made had a correlation coefficient of 0.47. Also the reason that each component of the metric was weighted how it was is because if how high the correlation coefficient was. RPI had the lowest correlation, so it was weighted the