gabednconfused
That/Dude
- Joined
- Nov 7, 2011
- Messages
- 2,554
I was thinking at lunch today about that quote of Richt going 'off feel' with the QB benching of Kosi, the onside kick, even some of his play calling...
So I ended up asking myself "I wonder if our team has an in-game simulation model?"
I know it sounds daft to some, but in this day in age, you need data. So I did some googling and found this company that seems to build an in-game computer model for CFB teams. Seems either Bama or Clemson use the software to supplement coaching decision making in critical situations (I suspect its Bama, based on the Kiffin tie-in). These situations are based usually on points differential, down and distance, and clock time left. They are influenced by statistics (e.g. moneyball type historical data).
I came across this company Championship Analytics, which seem to have gone to the absurd length of patenting their tech to do this: https://patentimages.storage.googleapis.com/99/f5/a0/b5fcdb1e86fcdd/US9424615.pdf
Personally, I woulda just went to the computer science dept in McArthur and found the dean to create a Phd project for 5 students to come up with my own model. I don't know the fidelity of the Championship model, so in cases like this I would say make our own. Also, its cheaper.
We can think of two at least three poor strategic decisions from the past game vs UVA which probably negatively effected our win probability. There are probably a host of other situations which can be influenced by it.
The thing with in-game statistical models is, as you play you re-run them constantly, changing the analysis from predicted to actual, basically making the accuracy better. Over the length of a game, that might be worth 1 score, or maybe just a few yards or TOP, but raising win probability from 45-51% is all you need some times (just an example).
------------------
Reason I am so passionate about this- I work for a huge sports team as an engineer (in a sport where engineering matters) and we develop our own strategy computer models to help get us to victory lane. All of our competitors do. Our computer models directly effect the outcome of our races, and what we do during the races, and they change as things progress as well, opening up new possibilities to exploit our competition.
This is very applicable to football, and seemingly isnt being utilized on a wide-spread scale.
Hope there are coaches reading this board... time to build a computer model guys. Gimmie a call for some help.
So I ended up asking myself "I wonder if our team has an in-game simulation model?"
I know it sounds daft to some, but in this day in age, you need data. So I did some googling and found this company that seems to build an in-game computer model for CFB teams. Seems either Bama or Clemson use the software to supplement coaching decision making in critical situations (I suspect its Bama, based on the Kiffin tie-in). These situations are based usually on points differential, down and distance, and clock time left. They are influenced by statistics (e.g. moneyball type historical data).
I came across this company Championship Analytics, which seem to have gone to the absurd length of patenting their tech to do this: https://patentimages.storage.googleapis.com/99/f5/a0/b5fcdb1e86fcdd/US9424615.pdf
Personally, I woulda just went to the computer science dept in McArthur and found the dean to create a Phd project for 5 students to come up with my own model. I don't know the fidelity of the Championship model, so in cases like this I would say make our own. Also, its cheaper.
We can think of two at least three poor strategic decisions from the past game vs UVA which probably negatively effected our win probability. There are probably a host of other situations which can be influenced by it.
The thing with in-game statistical models is, as you play you re-run them constantly, changing the analysis from predicted to actual, basically making the accuracy better. Over the length of a game, that might be worth 1 score, or maybe just a few yards or TOP, but raising win probability from 45-51% is all you need some times (just an example).
------------------
Reason I am so passionate about this- I work for a huge sports team as an engineer (in a sport where engineering matters) and we develop our own strategy computer models to help get us to victory lane. All of our competitors do. Our computer models directly effect the outcome of our races, and what we do during the races, and they change as things progress as well, opening up new possibilities to exploit our competition.
This is very applicable to football, and seemingly isnt being utilized on a wide-spread scale.
Hope there are coaches reading this board... time to build a computer model guys. Gimmie a call for some help.
