Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
GPS Applications in Team Sports
Transcript of GPS Applications in Team Sports
How do we collect the data
How do we store the data
How do we process the data (tagging, summarizing, aggregating...)
How do we integrate the data
How do we analyze and visualize the data
How do we communicate the data
Garbage in, garbage out
Why are we collecting the data
What data do we want to collect
How is data affecting vital behaviors and decision making
How is data being used
Descriptive or prescriptive
How to provide meaning to the data
How to avoid paralysis by analysis
Create database of drills
Group drills based similarities
Find optimal number of groups
Check at what GPS metrics do they differ
Don't need to buy GPS units - borrow!
More ART than a science
What GPS metrics to consider?
How to aggregate individual performance into drill performance?
What method of clustering to use?
Similar to drill classification, but this time we classify GPS metrics
Latent Variable Modeling
The goal is to find groups of similar GPS variables - are they measuring the "same" latent quality?
Reduce number of variables
Combine with drills classification to see what quality is overloaded with what drill
New food on the table
Only physical data - better if we combine with technical
PRESCRIPTIVE - what to do?
Drills/metrics are not that "separable" ( a lot of overlap)
Is there a (physical) performance difference between positions [game demand]?
Is there a difference in physical qualities between positions?
Mind your metrics - garbage in, garbage out
In what metrics (or latent variables) do positions differ?
Are the drills we do "overloading" these latent variables?
Do drills performance differ based on position played?
Do drills performance differ based on individual qualities?
Do our drills providing enough of overload based on position played and physical qualities?
Where do they provide overload?
What can we do to
DO NOT focus only on physical!
Worst case scenarios
- "provide for the worst, the best/average can take care of itself" (Nassim Taleb)
Velocity, acceleration, MP
Low intensity, high intensity....
Who decides on the cut-off?
Should we avoid the bins?
BikeScore(tm) applied to velocity/acceleration/ load
Should be apply more "data mining" to find the best approach? No free lunch
What is the
Absolute zones better when comparing individuals (e.g. positional differences), relative takes individual qualities into consideration: better at evaluating "load" on the body
Relative zones (based on some "important" physiological threshold)
How do we know it is important? Why is "load" in that zone important? Modeling?
E.g. MSS, MAS, CV/CP, MLSS...
Problem of the "step" function
With any metric it is important to have a context
Game is the toughest load
Use it to provide context
How "hard" is the load compared to the game? (ind. game perf.)
Normalized week load: in games played
Different minutes played problem!
Acute training load
Chronic training load
TSB = difference
Impulse -> Response
How are training loads affecting adaptation
More individualize approach
Might help find out what metrics are important (validate the metrics)
Response = change in individual qualities, change in game performance
As with Banister's Impulse Response model we try to predict performance
Can we predict injuries?
Supervised vs. unsupervised learning
You must have injuries to be able to predict injuries - ethical?
Group models or individual models?
"Fragility is quite measurable, risk not so at all, particularly risk associated with rare events"
"It is far easier to figure out if something is fragile than to predict the occurrence of an event that may harm it"
We collect so much data almost every day - why do we still need to test?
Assuming validity, reliability and sensitivity of the metrics
Rule of big numbers -> constraints
Might be very useful (why?)
Use MP approach (validate first)
Test 3 shuttles of 30-60s, 5min and 10min
OR: Use "testing without testing"
Similar to "force-velocity" gives you energetic profile of the player
Can be used in calculating load and....