Current Agenda (by Sam)
1. Team Exploratory Data Analyis: Find strongest basic correlation between basic stats and winning %
2(Concurrent with 1): Take Current 2025 25-man rosters and data scrape, Research best ways to predict player performance and test out with 2023-2024 data. This will most likely be challenging. 
Look into SIERA, ERA+, FIP, xWOBA, Barrel%, etc 
3...........



Original Google Doc:
Priority Rankings
2025 Standings Predictions using Sam’s idea
Breakout Candidates + Rookie Predictions 
Cost per WAR analysis and finding model to predict dollar-per-win
What wins championships more often: Pitching versus Hitting 
Room for other if time 



Miraj - I’m a data science major. Yankees fan. I have decent comfortability in python and R. 
Sam - Data Science, Cubs fan
Adam - Applied Stats, Red Sox fan, like making spreadsheets and graphs
Avi- Data Analytics, Nationals fan 
Tyler - Business Analytics, Pirates fan, Comfortable in Python, semi-comfortable in R
Angad - Data Analytics, comfortable in statistics 



Ideas for project


Top 5 breakout candidates for hitting and pitching 
Standings predictions 
MVP candidates predictions 
CY Young Predictions 
Win/loss prediction
Stat prediction for players/team
Gold glove predictions
HOF predictions
Difference in value of oWar and dWar
Taking predictive stats, and combining with proper result stats to estimate win counts for every team. 
By sunday at 1 flesh out your idea




Sam’s plan (cuz i wont be there)
Either using Lahman package in R or bref/fangraphs, scrape historical data into csv/excel/dataset
Split randomly into 80/20 train/test set. 
Find correlation between teamwide stats and wins using plots and visualization
Using stats with highest correlation, test linear / multiple regression models with test set, see which one is most accurate. Also look into and research neural network (go to machine learning club!)
Now that we know which team stats are most predictive, we now need to predict team/individual stats for everyone. By this time hopefully most/all free agents have signed (mid feb?) Repeat process with projected 25 man rosters with individual stats (This can also be done in conjunction with parts 1-4 by splitting up the group
Total up team stats
Using our most accurate model, project 2025 standings based on team stats we predicted and see if they make sense. 


Miraj:
Obtain data sets from CSV’s or baseballR scraping. 
Want to look at all statcast bat to ball, exit velos, pulled fb rate (all of them on savant honestly) from 2015-2024 and get the strongest correlators to wOBA and xwOBA. 
From here we need to get something with around 90-95% confidence that can predict this. 
From this, parse the top 100 hitters and find any that are not usually up in the top hitters. 
Most likely we want to lower the PA to 200 PAs
Using this data, I also want to predict HR data and see what correlates to the most homeruns.
ANYTHING PITCHING:
I want to search through the stuff+ data and see if theres anything I like from pitchers
From here I want to do heatmap analysis and see where each pitch gets hit and see if there is any disparity from heatmap to stuff+ and find if we can fix a few pitchers. Need to find abnormally high stuff+ but high FIP (4.00+)

Rookie Predictions 
Cost per WAR
Angad:

Use data from baseball reference and baseball savant and other sources to find the WAR and salaries of the top players at each position (pitchers, infielders, outfielders). Using the data we can visualize Salary vs War to find out which players are “overpaid” compared the the average at their positon. We can then predict based on salaries for next szn and statistics from a players career (or just last season) which will perform below expectation for their salary. 

I don’t know too much about baseball but this is just a general idea I had.