- Joined
- Oct 12, 2011
- Messages
- 12,732
When Miami hired Manny there was talk about him being into analytics. I tend the cringe when I hear that someone is a "data guy" or "into analytics" because it often means they enjoy reading stats and probably don't understand optimization. I don't work in football or sports for that matter, and I don't have a Ph.D. in Statistics, Data Science or Data Analytics but I generally understand optimization (especially within supply chain).
All of the above considered if I were UM I would use analytics to optimally select an offensive system and coach. By selecting either an offensive system or coach first you would be potentially introducing a massive constraint. You can't look at just stats when hiring a coach because that assumes all things are equal between UM and the targeted coaches school. It is likely that whoever Miami hires would have different advantages at Miami then their other school and you would want a model that doesn't select a coach that projects to do worse at Miami.
As I previously said, I don't have experience in sports management and the below is meant to be general. Someone that worked in football analytics for ten or twenty years would likely come up with something a lot better.
Here is how I would approach an optimization model
Goal: Maximize net points per game by selecting an offensive system and offensive coordinator
Constraint 1
The offensive staff must fit within the budget UM is willing to pay
Constraint 2
The model would need to consider the type of players UM has been landing over the last several years
The optimization model would need to select the system that generates the most net points per game while being affordable by Miami. Looking at the current roster and recruiting trends and comparing it to every other team in the FBS it would align us with the offensive that has the highest correlation to ours and one that we could afford to pay a staff. The current bias (mine included) is that Miami should be running a spread, but the model would need to solve for what variation. Without using some type of statistical approach it is just a guess on how well Yost, Long, or any other OC would do here. Using PPG, YPG should be variables and not deciding factors. Though at the very least we can be confident that anyone would probably do better than Enos.
All of the above considered if I were UM I would use analytics to optimally select an offensive system and coach. By selecting either an offensive system or coach first you would be potentially introducing a massive constraint. You can't look at just stats when hiring a coach because that assumes all things are equal between UM and the targeted coaches school. It is likely that whoever Miami hires would have different advantages at Miami then their other school and you would want a model that doesn't select a coach that projects to do worse at Miami.
As I previously said, I don't have experience in sports management and the below is meant to be general. Someone that worked in football analytics for ten or twenty years would likely come up with something a lot better.
Here is how I would approach an optimization model
Goal: Maximize net points per game by selecting an offensive system and offensive coordinator
Constraint 1
The offensive staff must fit within the budget UM is willing to pay
Constraint 2
The model would need to consider the type of players UM has been landing over the last several years
The optimization model would need to select the system that generates the most net points per game while being affordable by Miami. Looking at the current roster and recruiting trends and comparing it to every other team in the FBS it would align us with the offensive that has the highest correlation to ours and one that we could afford to pay a staff. The current bias (mine included) is that Miami should be running a spread, but the model would need to solve for what variation. Without using some type of statistical approach it is just a guess on how well Yost, Long, or any other OC would do here. Using PPG, YPG should be variables and not deciding factors. Though at the very least we can be confident that anyone would probably do better than Enos.
