[STATSBOMB] Weird freeze frame player Ids #474
Description
Because of Issue #464 I figured I'd give it a try to convert StatsBomb 360 data into a TrackingDataset.
It does not seem to be super complicated, but I'm unable to convert this TrackingDataset to a dataframe.
Below is a minimal working example.
from kloppy import statsbomb
from kloppy.domain import Orientation, TrackingDataset
match_id = 3893829
statsbomb_dataset = statsbomb.load(
event_data=f"https://raw.githubusercontent.com/statsbomb/open-data/master/data/events/{match_id}.json",
lineup_data=f"https://raw.githubusercontent.com/statsbomb/open-data/master/data/lineups/{match_id}.json",
three_sixty_data=f"https://raw.githubusercontent.com/statsbomb/open-data/master/data/three-sixty/{match_id}.json"
).transform(to_orientation=Orientation.STATIC_HOME_AWAY)
dataset = TrackingDataset.from_dataset(
statsbomb_dataset.filter(
lambda event: event.freeze_frame is not None
),
lambda event: event.freeze_frame
)If we now run
dataset.to_df()We get
File ~/kloppy/domain/models/common.py:1274, in Dataset.to_dict(self, orient, *columns, **named_columns)
[1272](1272) items = defaultdict(lambda: [None] * c)
[1273](1273) for i, record in enumerate(self.records):
-> [1274](1274) item = transformer(record)
[1275](1275) for k, v in item.items():
[1276](1276) items[k][i] = v
File ~/kloppy/domain/services/transformers/data_record.py:76, in DataRecordToDictTransformer.__call__(self, data_record)
[75](:75) def __call__(self, data_record: T) -> Dict[str, Any]:
---> [76](:76) return self.converter(data_record)
File ~/kloppy/domain/services/transformers/attribute.py:283, in DefaultFrameTransformer.__call__(self, frame)
[279](:279) for player, player_data in frame.players_data.items():
[280](:280) print(player, player_data)
[281](:281) row.update(
[282](:282) {
--> [283](:283) f"{player.player_id}_x": player_data.coordinates.x
[284](:284) if player_data.coordinates
[285](:285) else None,
[286](:286) f"{player.player_id}_y": player_data.coordinates.y
[287](:287) if player_data.coordinates
[288](:288) else None,
[289](:289) f"{player.player_id}_d": player_data.distance,
[290](:290) f"{player.player_id}_s": player_data.speed,
[291](:291) }
[292](:292) )
[294](:294) if player_data.other_data:
[295](:295) for name, value in player_data.other_data.items():
AttributeError: 'NoneType' object has no attribute 'player_id'This seems to happen because the Player Ids of the freeze frame players are not the same as the actual player Ids in the meta data. For example:
Player(player_id='T16803-Ef981129f-d390-4875-921c-e3c08203c2fd-3', ...),
Player(player_id='T16803-Ef981129f-d390-4875-921c-e3c08203c2fd-4', ...),
Player(player_id='10272', ...),
Player(player_id='T16803-Ef981129f-d390-4875-921c-e3c08203c2fd-6', ...)The only "correct" player_id is 10272 which I assume belongs to the player executing the event.
FWIW, I would love to support StatsBomb freeze frames in unravelsports, but not sure that’s doable this way.
We should probably create a to_df("rows") way to combat this.