I've begun the model fitting for the semester project in the Statistical Analysis course I'm taking--I'm trying to predict the number of points the San Francisco 49ers scored in each game over the past 2 seasons based on various game statistics not including the number of touchdowns, field goals, etc. I'm trying to predict scores based on data like rushing yards, number of offensive plays, turnovers, etc.
I used to do this as a hobby in my younger, pre-fatherhood days when I was glued to the television all weekend during football season. On Monday I would look at game stats and try to predict the outcome of games. A lot of it was gut feel, but one standard rule of thumb was this: each turnover is worth 4 points to the opposing team.
In the first model that I've fit, the parameter estimate for opponent turnovers is 3.55. That means that each opponent turnover was worth 3.55 points for the 49ers.
My rule of thumb seems to be justified!
Just for giggles, I fit the model to the data from this past Super Bowl. The Niners scored 31 points in that game, the model predicts 29.5.