- Title, Authors: yes
- Motivation: Can you add some information about the value ($ billions?) of the industry? The team mentioned the problem is not trivial (nice)
- Data: size of data identified, a look at the columns provided. Would be nice to identify on this slide what is your outcome as well (make it clear, a point of its own: total points is our outcome). Seasons starting in 2016. Why did the number of records increased over time (seasons)? What is clean sheets again? For your outcome, total points is for each player or for each team? Spell out player positions (I am not familiar with them).
- Modeling: LR (as baseline). RF or DT, FFNN - looks good
- Metrics: move the explanation on how you split the data to the Data section. I didn’t understand what you mean by random split of your data using K-fold (it has a time dimensi you wouldn’t want to do this). RMSE and MAPE (good!).
- Would be cool to see more variables added, not sure if there were any sub questions alongside the main question was but would be interesting
- Very detailed EDA! the correlation matrix is something that we might borrow for our own addtl. team EDA :)
- n/a
- I thought it went well
- looks good
- Good EDA.
- Would it be possible to have an EDA on players and their average stats for that? Knowing that the point is to draft the best players, I think it would be great context to have so people would know what is typical stat-line. Otherwise, I thought the EDA and the modeling looks robust and is pretty good!
- N/A
- t-test for comparing to baseline seems like an effective approach to validate model usefulness
- The score data is rolled up weekly, but it could be interesting to have different rolling periods for evaluation; like what if a player had a minor injury and experienced a lower scoring potential for a number of weeks.
- Have you considered LSTM? I'm wondering because of the time series aspect whether this might be a useful model architecture
- Plots were very helpful in illustrating the arguments/hypothesis being made. Also thought the piece on the context/motivation was vey well presented. Modeling and evaluation metrics sounded reasonable! Would there be any feature that captures how the inputs change as the season progresses?
- Would be helpful to maybe do a practice run together before class
- I think it went really well. Very interesting project.
- I don't think so, thought it was great! Maybe in the final presentation include a few more visual aids. I think some further explanation of some of the football terminology used in the data set would be helpful as well!
- I know this is ticky tacka, but the enthusiasm did not translate to me, and it felt very pedestrian.
- some diagrams to explain the parameters will help
- Try to use less technical terms and focus on the most important aspects of the presentation.
- Overall, I really enjoyed the presentation and thought the slides were easy to digest. I think the only thing that would be helpful to elaborate on is how fantasy premier league works (especially as someone who doesn't play) and what are typically more important metrics to have a good team
- I thought your plan was well thought and and communicated very clearly!
- More visuals could be helpful
- The plots were easy to read and understand, and the slides were clean and organized
- Wasn't familiar with MAPE, potentially just describe what you're trying to achieve when selecting this performance metric (how does it benefit your dataset)
- Well-structured and great presentation!
- nice job, very interesting concept
- You guys did really good- very detailed
- Looking forward to seeing the outcome of this!
- I thought the initial idea was splendid
- good luck
- Overall good presentation.
- Can't wait to see how your model performs as the season goes on
- Nice work!
- Great clear presentation and flow
- hopefully your model will have good predictive capability and will help you make some money on sports betting!
- Awesome job!
- N/A!