Reply To: Work Load and Injury

Mladen Jovanovic

10/06/2016 at 20:17 #18094

Hi Doc,

Thanks for the post. I am developing one injury prediction model using GPS metrics, RPE and Wellness scores – although predictions are perfect for “training dataset”, the predictions for the “hold out” data set are not great (AUC 0.7). But again it could be due small data set (one season is training data set, and another season is holdout).

Generally, we are juggling with models and assumptions. These needs to be validated with the real supervised model. The problem is that there are a LOT of factors involved and we might be dealing with ignorance problem.

I guess Gabbett model (actually Banister) of checking the ATL and CTL is a generally good/useful strategy, but we still don’t have a proof it works, even if Gabbett et al. published studies on it, they haven’t used holdout data set.

Hope this helps,

Mladen