Big thanks to Timothy Wynn who recently brought back a post with a great list of resources. I was wondering how people use this information... Have you created your own rating system starting with raw stats or do you use one of the many power-rating type systems already out there? Or maybe some combination?
I first got interested in sports-gambling when I was creating an NFL power-rating system from raw game data. After I finished it I noticed that the prediction outcomes were very similar to prediction outcomes I could generate using freely available online ratings. It is much easier to use these pre-rolled ratings than rolling my own so I decided to use these easily found online ratings to test NFL prediction theories.
After that I did the same thing with sports other than the NFL.
I have been working on prediction systems for MLB, NBA, NCAA FBS, FCS, and NCAA BB using freely available online ratings. I have learned a lot doing this and one conclusion I have reached is that the accuracy of a particular rating system is less important than understanding how that system relates to information contained in the lines coming out of Vegas. Or maybe more accurately stated, how a system relates to the information that is found between the lines coming out of Vegas. To put it another way, working out how a rating system connects with winning (not win %, but winz $) is more important than having a rating system that gets every detail correct.
I start out my experimenting by picking a rating system and testing basic ideas (theories) against historical data. Usually a pattern starts to develop that leads to refining and retesting. If something works for an entire season, the next step is to work it over multiple seasons of data. For example, I think strength of schedule has a strong role in prediction success . So I searched online for a variety of SOS ratings and started testing. Once i found a rating that worked for me I incorporated that into the prediction model.
Edited 9/7/21 at 4:41PM by jakesjewels - No reason listed.