ZRod
Heisman Trophy Winner
Ok. What they can do is keep track of all of these hundreds of variables for every player (or even a sample of players) for every practice. And also keep track of which players were injured during that time to use as a response variable.
Then they can attempt to build a model to predict injuries based off of a subset of the data, and apply it to a separate subset of the data. E.g. a simple model could have 2 predictors (but probably wouldn’t work well) - in the 3 practices leading up to injuries, the players were at max speed more than 50% of the time, and changed direction over 60 times, and players that weren’t injured were usually below at least one of those amounts.
If they apply it to the separate subset and their model correctly picks out 20% of the injuries the day before they happened, they can then use the model in real time. So any player that has the values that put them over that threshhold according to the model can be held out of practice the next day.
They would just need to decide what amount of accuracy is worth the risk of keeping a player out of practice. If it’s only correct 5% of the time they might say screw it and not use it at all. If it’s correct 50% of the time it’s a no brainer to use it. Again, I don’t know if they’re doing this but it’s something they could try with that data.
They might also simply be making sure the players are honest with them on how they feel.
Must. Contain. Nerd bulge.
This is pretty cool. It's crazy that we have this technology at our disposal now. That being said it seems like there are a lot of variables that would affect the accuracy of their models, but the amount of exertion on the body is probably the biggest one and they can measure that. If you could link this with sleep, diet, and tie in flexibility/muscle imbalances you would have the ultimate system.