This repository contains an against-the-spread (ATS) predictor for college football games. There are two sources of data:
- Game-by-game stats are from the SportsReference API
- Historical betting lines and odds are from sportsbookreviewsonline.com
There are currently five code files in the repository:
- ats_picks.py: Generates picks and bets for upcoming games.
- ats_predictor.py: Looks for the best model and betting strategy by running simulations on past games.
- cleaner.py: Cleans the historical odds spreadsheet into another spreadsheet that is easier to work with.
- create_input_dfs.py: Takes the data from the two sources to create input stats for each game.
- fix_names.py: Contains a dictionary which is used to convert names in cleaner.py.
I think that, at the moment, the model doesn't really show any significant ability to predict against-the-spread. Thus, I predict that as the number of games goes to infinity, the model will approach 50% ATS. This obviously does not really help.
I think that the scores are predicted decently well, but as far as making money is concerned I'm not there yet. Luckily, collegefootballdata.com also has an API that has drive and play-by-play data (along with everything else I could want), so that's where I'll start when I resume this project.
There's also the fact that making money is a lot harder than it could be because traditional sportsbooks take such a big cut. So, I think it doesn't make as much sense to resume the project until decentralized betting markets are a bit further along. Projects like Augur are what would make it really cool to start building with this project again. The fees will be a lot lower and so it will be a lot easier to be profitable, there will be no limits (if I would ever get that far, haha), and the best part is that I could integrate my work with with their APIs to automate getting lines, making bets, etc.
- Adjust game stats for opponent
- I actually want to make picks using the opening spread, not the closing one
- A moneyline betting strategy