In this video I wanted to demonstrate how one could use differing team stats to be able to create their own power rankings.
In this video, the stats used, and the outcomes will not be the same as yours, as this was a quick demonstration, and not meant to be the answer to the perfect model. Now, doing this can work for just about any sport where there is a spread, whether it be NHL, NFL, Soccer, NBA, etc.
if you can be the sport with a spread, this can help.
The basic steps are to find stats that you believe will predict a score, create a linear regression for those stats. Then project out the entire season of values, and build a power ranking from them.
When you are building a power ranking for the predictions, you will follow the standard methods of creating a power ranking by finding the home margin of victory (Positive or Negative),and SSE. Use solver, and away you go!
