search

PySport / kloppy

kloppy: standardizing soccer tracking- and event data
https://kloppy.pysport.org
BSD 3-Clause "New" or "Revised" License
326 stars 55 forks source link

Track player position changes #307

Closed DriesDeprest closed 3 days ago

DriesDeprest commented 3 months ago

Created this draft PR related to: https://github.com/PySport/kloppy/issues/277, to get initial feedback on the implementation prior to starting updating all the deserializers. This is also based on the discussion made here: https://github.com/probberechts/kloppy/pull/1.

As a sake of example, I've started with updating the Statsomb deserializer to support the Player.positions.

@JanVanHaaren @koenvo happy to hear what you guys think.

DriesDeprest commented 2 months ago

@koenvo @JanVanHaaren did you guys find the time to take a look this PR? I'd like to continue the effort, but first want to check whether you agree with the implementation proposal.

JanVanHaaren commented 2 months ago

Thank you for your proposal, @DriesDeprest! Your approach generally looks good to me.

The following thoughts crossed my mind while looking through the code.

  1. I see similarities with the formation changes for teams. Would the position changes be used in a similar way?
  2. Would it be useful to know the end timestamp for each position in addition to the start timestamp?
  3. Are the timestamps ever in the mm:ss format instead of the HH:mm:ss format?

What are your thoughts, @probberechts?

DriesDeprest commented 2 months ago

Thanks for reviewing @JanVanHaaren!

1.

I noticed the same similarities and raised it here: https://github.com/PySport/kloppy/issues/301. How I see it, you have two options on how to get the actual formation / position of a team / player for each event. a) You create a specific event (FormationChange/PositionChange) for it and using the state builder, you can get the state formation/position at time of execution of the event. b) You have a formation/position_change attribute in the Team/Player object, instead of a formation/position attribute, and with the right method position/formation() you can get the actual formation / position of a team / player for each event.

For formations, we have historically chosen to go for implementation a) but for positions we are now building implementation b).

The reason why I prefer option b) over option a) is because it also offers you a standardized data format for storing formation changes of a team and position changes of players of a football game across vendors, besides the ability to get the actual formation / position of a team / player for each event

  1. See https://github.com/PySport/kloppy/issues/277#issuecomment-2015474935 for why we don't store the end timestamp of a position change.

  2. I believe so, but will double-check once I continue the implementation, after we agree on the high level implementation.

DriesDeprest commented 3 days ago

Closing this PR, since we added position changes with: https://github.com/PySport/kloppy/pull/326.