Sports betting is one of these perfect problems for machine learning algorithms and specifically classification neural networks. Tons of data available and a clear objective of picking the winner!
For NFL, you could use sites like footballdb.com to fetch historical data for your prediction model. Once the data is fetched and categorized into logical groups and value scored, you could use ML.NET to applying Machine Learning and create prediction models to these datasets can yield predictable results.
What you can do then, is to take some sample data and back-test your model and see how well your prediction model perform. Tweak your model until you get a more consistent result. Finally, you can test it again a current and future games. Albeit, it wouldn't give you a 100% accuracy but at least will put you in the ball park.
Hi tamasnorby - send me an email at hello@xamotoolkit.com.
Interested,how can we talk?
I will carve out some time to finalize my prediction models before the football season starts. As a football fan, this one is my favorite on the list.