Multi-Mechanical Modelling using Wearable Data
I had posted this thread on my twitter account, but also wanted to post it here as well.
_____________________________________________________________________________
I've had some questions since the
@catapultsports hockey webinar, and thought I would share the concept behind how I report daily volume and intensity through a combination of metrics.
I built an equation to report through the concept of "Multi-Mechanical Modelling", which I heard about on
@strengthofsci podcast #259 with @SharkeyStories . The premise is that volume and intensity are not a singular metric but is best represented by combining multiple metrics.
Taking the "MMM" concept, I wanted to apply both my experience alongside some statistical concepts - so performed a Principal Component Analysis on game data to determine which were the most valuable metrics to include in my model.
From here I was able to extract the valuable metrics that best explained the variance within the dataset. I did this for both volume metrics and intensity metrics. You can use a scree plot to help decide how many features to include.
Once I determined the metrics I wanted to include, I was able to use the results of the PCA to assign weighting factors to each variable within the model. I use the average for the day divided by the game average of the same variable.
The resulting equations (and the PCA results) allowed me to drop many of the possible metrics that I could report on as well as apply the relative importance of each variable.
All in all, this gives me a number I am confident in that best represents the daily volume and intensity of the session in a way that is easily digestible by the coaching staff (both as a team average and for each position).

0
Please sign in to leave a comment.
Comments
0 comments