I am trying to use math to predict NFL fantasy football scores. My current process for projecting a players score is as follows:
For every team (32 teams), I list the average points it gives up to opponents by position (K, DST, QB, WR, RB, TE). So I have 6 sets of 32 numbers.
For every position, I average its respective set of numbers and subtract the mean from all 32 data points. This leaves me with 32 "Points to add" numbers for each position.
I then add that number to each player fitting in the category. Obviously football cannot be 100% determined with numbers, but I am wondering if my theory at least makes any sense. I feel like I should be taking standard deviations or variances into account?
Here's an example of my process for one position:
Let's say there are 4 NFL teams: Pats, Broncos, Saints, Cowboys.
The Pats average 10 points per game to the position QB. The Broncos, Saints, Cowboys average 12, 14, and 18 respectively. The average points per game given up to that position is then 13.5. (average of 10,12,14,18)
Subtracting 13.5 points from the Pats average of 10 points means we expect the Pats to give up 3.5 LESS points on average to opposing QBs. Similarly we would expect the Cowboys to give up 4.5 MORE points to opposing QBs (18-13.5).
So I then find the QB playing the Pats and subtract 3.5 points from his mean. Likewise I find the QB playing the cowboys and add 4.5 points to his mean to get a projection.
Again, I understand nothing is absolute, but can this be modeled better with math, like maybe I shouldn't be subtracting from the average but some other number derived mathematically? 1 standard deviation away? Just an interesting question for those who understand probability and non ideal distributions better than me.