So I’ve been working on this project for a few years with mixed results, this year I started looking at things a bit different. Without divulging my formula its basically ranking each team using weekly stats to determine their grouping (Group 1 = top 4 teams; Group 8 = bottom 4 teams) from there I use a forecast formula to project the scores of future games based on the previous weeks stats. The results vary from week to week with significant drops in percentage wins in weeks 7 and 11. Still working on why it seems to be those two weeks as to why the formula is creating those dips in those weeks as the other weeks typically generate 53% or better.
Below are some color coded charts that illustrate the average Vegas spread based on the visiting team and home team grouping. When running a confidence formula to determine the low and high Vegas spread, I’ve discovered that when the actual Vegas Spread is within the low/ high values I’m getting a better win percentage against the spread.
The charts below are broken down respectfully (Average, Divisional, Conference, and Non-Conference) showing that the Vegas spread changes based on the type of game.
https://drive.google.com/file/d/0B-L...it?usp=sharing
The next image is of my win percentage which is calculated over a four year span 2009 to 2012 or 960 regular season games. Now the current set up is formula based selecting 45% of those games however it’s picking a high win percentage. When I include the variable above its selecting roughly 2% of the 960 games.
https://drive.google.com/file/d/0B-L...it?usp=sharing
My question is if having a better understanding of what the Vegas spread should be based on the match up grouping, I’m thinking that I could use that to adjust the projected scores to either add or subtract points based on the difference of the actual Vegas Spread and the projected Vegas Spread. Does this logic seem right or am I on the wrong path?
I’m hoping I made some sense out of all of this but its late and I’ve been working on this stuff tonight way too long.
If you have questions please ask.
Thanks
ny92mike