Peer Review #2 (ss3759)

[saaqebs]

This project's goals aren't clearly outlined from the midterm report; however, it can be inferred both from the project's name as well as their preliminary analysis that they are trying to compile and analyze what would make up the best (or dream) team for fantasy football. This sounds super cool! I bet Patrick Mahomes will clearly be the quarterback for this dream team.

Things I enjoyed about this proposal:

• The dataset is clear-cut and straightforward; it appears very easy to use and analyze based on the depth of the preliminary analysis.
• The use of feature engineering to hyper predict the rush/pass/receiving yards is extremely useful to see trends amongst when players may peak during a season.
• The correlation of features was so colorful and beautiful to see implemented to see which features may be correlated with one another. Our team should try to implement something similar to visualize similarities!

The three things that I am concerned about is:

• The way feature engineering is being implemented simply looks at the previous week, implying that the "hot streak" would only look back a week. Another feature that may be helpful is to see the season's cumulative stats at that current week.
• The datasets are both full of information, but I fear that the statistics brought from the fantasy football dataset lack the proper insights compared to the raw statistics. Likewise, if the goal is to focus on optimizing fantasy football points, then the raw data would need to be converted into projected fantasy points based on the proper equations.
• I am not sure how play-by-play data would create a deeper insights into the project, especially when focusing on fantasy points. I am curious as to how that would be thoroughly implemented!