Skip to content

Excel and MATLAB files to analyse the team colors of the NCAA

Notifications You must be signed in to change notification settings

karsonkevin2/NCAA-Colors

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NCAA-Colors

Excel and MATLAB files to analyse the team colors of the NCAA

Excel File

This file has data for all NCAA Division 1 teams. The information was compiled secondhand from a variety of aggregation sites listed in the first sheet of the Excel file. Where gaps existed in the data, I went to the University Brand websites to find the missing information.

Colors are presented using the 8 bit RBG color scheme defined from 0-255. Each teams primary/secondary and accent colors are included as both RGB values and with colored boxes generated using VBA.

Towards the right in the NCAA v2 sheet are input fields. You can put in colors for a real or made up team. For example, input data for your highschool. Then you can sort the Min column to find the most similar college teams. The similarity metric used is presented in the whitepaper here, https://www.compuphase.com/cmetric.htm

Colors similar to Iowa

Similar to Iowa

The colorRef sheet presents counts of colors grouped together. This uses a naive approach of fixed boundaries. This approach works well for computer vision, but human eyes do not work in the same manner. Some colors which may be similar to a human eye may seem different to a computer and vice-versa.

Naive Color Grouping

Naive color grouping

MATLAB File

The primary color data should be imported from the Excel document and labeled as variable aPrimary. The secondary color data should be imported from the Excel document and labeled as aSecondary.

With this data, run the file. The document was primarily for development and testing, so the code contains redundancies and minimal comments.

Figure 1 shows all of the primary and secondary colors and clustered colors using the human vision algorithm. Figure 2 shows the same data but with clustered data from the computer vision algorithm. Figure 3 shows how many data points belong to each clustered point. Figure 4 shows a pie chart of how common each color is. Figure 5 shows why equidistant color points will not work well for binning. Figure 6 shows the colors without any overlay.

Figure 1

Figure 1

Figure 4

Figure 4

Lines 5-7

%A = aPrimary;
%A = aSecondary;
A = [aPrimary ; aSecondary];

If you want to see data specifically for just primary or just secondary color change which lines are commented, such that only the desired data is uncommented.

Line 9

clusterP = 27;

This is the clustering for the data. 27 will be the number of clustered data points. Change this value smaller or larger to change the number of points. 27 was chosen for example since this is a 3x3 grid. Smaller and larger values will also work well and present interesting data.

About

Excel and MATLAB files to analyse the team colors of the NCAA

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages