raw_user_forecasts.csv a dataset of user predictions for the 2022-23 season from fivethirtyeight's "Elo Game."
"archive" is pulled from this kaggle dataset. It contains historical betting info.
This project was inspired by the elo system used by FiveThirtyEight. Elo is a score assigned to each team that represents their team-strength. It is derived from that team's performance in a closed pool of zero-sum games.
Using elo, you can calculate the odds of team A winning a given game as:
Here are the main features that go into updating elo scores in this system:
This is a parameter that controls how quickly elo updates according to recent games. A high K-factor indicates that recent games are more indicative of future success than older games, therefore increasing K will add recency bias to the rating system.
This is how different the outcome of the game is from what the Elo system predicted.