State¶

While working with event data it can be convenient to also use the current score. This can be used to determine if a team is winning, losing or drawing. A more generic name for the score is 'state'.

In this quickstart we'll look at how to use the add_state method of kloppy to add state to the events for later use.

Loading some statsbomb data¶

First we'll load Barcelona - Deportivo Alaves from the statsbomb open-data project.

In [31]:

from kloppy import statsbomb
from kloppy.domain import EventType

dataset = statsbomb.load_open_data()
print([team.name for team in dataset.metadata.teams])
print(dataset.events[0].state)

from kloppy import statsbomb from kloppy.domain import EventType dataset = statsbomb.load_open_data() print([team.name for team in dataset.metadata.teams]) print(dataset.events[0].state)

/Users/koen/Developer/Projects/PySport/kloppy/.venv/lib/python3.10/site-packages/kloppy-3.7.1-py3.10.egg/kloppy/_providers/statsbomb.py:67: UserWarning: 

You are about to use StatsBomb public data.
By using this data, you are agreeing to the user agreement. 
The user agreement can be found here: https://github.com/statsbomb/open-data/blob/master/LICENSE.pdf

  warnings.warn(

['Barcelona', 'Deportivo Alavés']
{}

Add state - score¶

kloppy contains some default state builders: score, lineup and sequence. Let's have a look at the score state builder.

In [32]:

dataset = dataset.add_state('score')

dataset.events[0].state

dataset = dataset.add_state('score') dataset.events[0].state

Out[32]:

{'score': Score(home=0, away=0)}

As you can see state is now filled with a score object. The object contains two attributes: home and away. Every event contains a score object which is automatically updated when a goal is scored.

Now lets have a look at how we can use the score state. First we filter on only shots.

In [33]:

dataset = dataset.filter(lambda event: event.event_type == EventType.SHOT)
shots = dataset.events
len(shots)

dataset = dataset.filter(lambda event: event.event_type == EventType.SHOT) shots = dataset.events len(shots)

Out[33]:

28

In [34]:

for shot in shots:
    print(shot.state['score'], shot.player.team, '-', shot.player, '-', shot.result)

for shot in shots: print(shot.state['score'], shot.player.team, '-', shot.player, '-', shot.result)

0-0 Barcelona - Lionel Andrés Messi Cuccittini - OFF_TARGET
0-0 Barcelona - Jordi Alba Ramos - OFF_TARGET
0-0 Barcelona - Lionel Andrés Messi Cuccittini - SAVED
0-0 Deportivo Alavés - Rubén Sobrino Pozuelo - OFF_TARGET
0-0 Barcelona - Luis Alberto Suárez Díaz - OFF_TARGET
0-0 Barcelona - Ousmane Dembélé - OFF_TARGET
0-0 Barcelona - Ivan Rakitić - OFF_TARGET
0-0 Barcelona - Lionel Andrés Messi Cuccittini - POST
0-0 Barcelona - Gerard Piqué Bernabéu - OFF_TARGET
0-0 Barcelona - Ousmane Dembélé - SAVED
0-0 Barcelona - Luis Alberto Suárez Díaz - OFF_TARGET
0-0 Barcelona - Ousmane Dembélé - OFF_TARGET
0-0 Barcelona - Jordi Alba Ramos - SAVED
0-0 Deportivo Alavés - Mubarak Wakaso - BLOCKED
0-0 Barcelona - Luis Alberto Suárez Díaz - BLOCKED
0-0 Barcelona - Philippe Coutinho Correia - BLOCKED
0-0 Barcelona - Jordi Alba Ramos - BLOCKED
0-0 Barcelona - Lionel Andrés Messi Cuccittini - OFF_TARGET
0-0 Barcelona - Philippe Coutinho Correia - BLOCKED
0-0 Barcelona - Lionel Andrés Messi Cuccittini - GOAL
1-0 Barcelona - Lionel Andrés Messi Cuccittini - POST
1-0 Barcelona - Ivan Rakitić - BLOCKED
1-0 Barcelona - Luis Alberto Suárez Díaz - SAVED
1-0 Deportivo Alavés - Adrián Marín Gómez - OFF_TARGET
1-0 Barcelona - Philippe Coutinho Correia - SAVED
1-0 Barcelona - Philippe Coutinho Correia - GOAL
2-0 Barcelona - Lionel Andrés Messi Cuccittini - SAVED
2-0 Barcelona - Lionel Andrés Messi Cuccittini - GOAL

In [35]:

dataframe = dataset.to_df(
    "*",
    home_score=lambda event: event.state['score'].home,
    away_score=lambda event: event.state['score'].away
)
dataframe

dataframe = dataset.to_df( "*", home_score=lambda event: event.state['score'].home, away_score=lambda event: event.state['score'].away ) dataframe

Out[35]:

	event_id	event_type	result	success	period_id	timestamp	end_timestamp	ball_state	ball_owning_team	team_id	player_id	coordinates_x	coordinates_y	end_coordinates_x	end_coordinates_y	body_part_type	home_score	away_score
0	65f16e50-7c5d-4293-b2fc-d20887a772f9	SHOT	OFF_TARGET	False	1	149.094	None	alive	217	217	5503	0.930417	0.645625	None	None	RIGHT_FOOT	0	0
1	b0f73423-3990-45ae-9dda-3512c2d1aff3	SHOT	OFF_TARGET	False	1	339.239	None	alive	217	217	5211	0.949583	0.336875	None	None	LEFT_FOOT	0	0
2	13b1ddab-d22e-43d9-bfe4-12632fea1a27	SHOT	SAVED	False	1	928.625	None	alive	217	217	5503	0.766250	0.430625	None	None	LEFT_FOOT	0	0
3	391bfb74-07a6-4afe-9568-02a9b23f5bd4	SHOT	OFF_TARGET	False	1	979.616	None	alive	206	206	6613	0.908750	0.483125	None	None	HEAD	0	0
4	5e55f5a5-954f-4cc4-ba6e-a9cf6d6e249e	SHOT	OFF_TARGET	False	1	1095.914	None	alive	217	217	5246	0.891250	0.311875	None	None	RIGHT_FOOT	0	0
5	1c0347cd-14dc-4aa8-91eb-520672a6cfe1	SHOT	OFF_TARGET	False	1	1842.287	None	alive	217	217	5477	0.900417	0.341875	None	None	LEFT_FOOT	0	0
6	7c3182af-c8a8-4c7c-934e-5c41c7b93c6a	SHOT	OFF_TARGET	False	1	2104.861	None	alive	217	217	5470	0.932917	0.545625	None	None	HEAD	0	0
7	39f231e5-0072-461c-beb0-a9bedb420f83	SHOT	POST	False	1	2248.168	None	alive	217	217	5503	0.807917	0.674375	None	None	LEFT_FOOT	0	0
8	062cdd08-8773-424f-8fc5-2e3d441c3c5c	SHOT	OFF_TARGET	False	1	2250.989	None	alive	217	217	5213	0.935417	0.516875	None	None	HEAD	0	0
9	c09e904d-6c8e-479d-af2e-c2c5863aca71	SHOT	SAVED	False	1	2308.083	None	alive	217	217	5477	0.853750	0.364375	None	None	RIGHT_FOOT	0	0
10	9eda2385-88bb-4c37-b039-a738c4cd8d4b	SHOT	OFF_TARGET	False	1	2434.592	None	alive	217	217	5246	0.910417	0.370625	None	None	RIGHT_FOOT	0	0
11	0b56df83-958e-4a31-9e5a-d741b4bf95cf	SHOT	OFF_TARGET	False	1	2610.612	None	alive	217	217	5477	0.887083	0.418125	None	None	RIGHT_FOOT	0	0
12	79851101-0f47-48e2-bdac-2de2b2bc89cb	SHOT	SAVED	False	2	159.524	None	alive	217	217	5211	0.973750	0.389375	None	None	LEFT_FOOT	0	0
13	d37ddbf4-05b6-4127-8ca6-287a069bd912	SHOT	BLOCKED	False	2	367.400	None	alive	206	206	6626	0.882917	0.286875	None	None	LEFT_FOOT	0	0
14	b8f9567a-b33c-4013-91a4-dd5c088dd14a	SHOT	BLOCKED	False	2	534.355	None	alive	217	217	5246	0.925417	0.300625	None	None	RIGHT_FOOT	0	0
15	2e2b7445-812c-4162-a5a2-4a99290359ae	SHOT	BLOCKED	False	2	596.388	None	alive	217	217	3501	0.786250	0.305625	None	None	RIGHT_FOOT	0	0
16	480a2132-0c0e-4a0b-95e7-eb994a1542ff	SHOT	BLOCKED	False	2	634.490	None	alive	217	217	5211	0.894583	0.323125	None	None	LEFT_FOOT	0	0
17	dfb55124-1e6a-475c-a460-3d9bf920bf04	SHOT	OFF_TARGET	False	2	740.847	None	alive	217	217	5503	0.833750	0.725625	None	None	LEFT_FOOT	0	0
18	ff334a7e-6ec2-4c71-9957-6d8accaf48df	SHOT	BLOCKED	False	2	936.156	None	alive	217	217	3501	0.834583	0.345625	None	None	RIGHT_FOOT	0	0
19	4c7c4ab1-6b9f-4504-a237-249c2e0c549f	SHOT	GOAL	True	2	1091.954	None	alive	217	217	5503	0.800417	0.563125	None	None	LEFT_FOOT	0	0
20	28175830-bbf0-4be9-baef-75c1de731809	SHOT	POST	False	2	1243.588	None	alive	217	217	5503	0.883750	0.588125	None	None	LEFT_FOOT	1	0
21	89cef214-dec8-43c7-8222-3ea8942ae868	SHOT	BLOCKED	False	2	1413.375	None	alive	217	217	5470	0.738750	0.534375	None	None	RIGHT_FOOT	1	0
22	0a78095f-dde1-4cba-8639-3ac2750d3a4d	SHOT	SAVED	False	2	1647.492	None	alive	217	217	5246	0.908750	0.375625	None	None	LEFT_FOOT	1	0
23	6e7d482d-30cc-4f5c-9b65-84c52ef8247c	SHOT	OFF_TARGET	False	2	1997.679	None	alive	206	206	6935	0.944583	0.418125	None	None	RIGHT_FOOT	1	0
24	5f8f624e-eae9-4e6b-a086-d864bc8f4140	SHOT	SAVED	False	2	2191.606	None	alive	217	217	3501	0.944583	0.388125	None	None	RIGHT_FOOT	1	0
25	683c6752-13bc-4892-94ed-22e1c938f1f7	SHOT	GOAL	True	2	2261.578	None	alive	217	217	3501	0.875417	0.388125	None	None	RIGHT_FOOT	1	0
26	252b3061-7d3f-4922-b04f-0b37a44c6300	SHOT	SAVED	False	2	2662.638	None	alive	217	217	5503	0.883750	0.575625	None	None	LEFT_FOOT	2	0
27	55d71847-9511-4417-aea9-6f415e279011	SHOT	GOAL	True	2	2802.770	None	alive	217	217	5503	0.932917	0.431875	None	None	LEFT_FOOT	2	0

Now filter the dataframe. We only want to see shots when we are winning by at least two goals difference.

In [36]:

dataframe[dataframe['home_score'] - dataframe['away_score'] >= 2]

dataframe[dataframe['home_score'] - dataframe['away_score'] >= 2]

Out[36]:

	event_id	event_type	result	success	period_id	timestamp	end_timestamp	ball_state	ball_owning_team	team_id	player_id	coordinates_x	coordinates_y	end_coordinates_x	end_coordinates_y	body_part_type	home_score	away_score
26	252b3061-7d3f-4922-b04f-0b37a44c6300	SHOT	SAVED	False	2	2662.638	None	alive	217	217	5503	0.883750	0.575625	None	None	LEFT_FOOT	2	0
27	55d71847-9511-4417-aea9-6f415e279011	SHOT	GOAL	True	2	2802.770	None	alive	217	217	5503	0.932917	0.431875	None	None	LEFT_FOOT	2	0

Add state - lineup¶

We are able to add more state. In this example we'll look at adding lineup state.

In [37]:

from kloppy import statsbomb
from kloppy.domain import EventType

dataset = statsbomb.load_open_data()

home_team, away_team = dataset.metadata.teams

from kloppy import statsbomb from kloppy.domain import EventType dataset = statsbomb.load_open_data() home_team, away_team = dataset.metadata.teams

/Users/koen/Developer/Projects/PySport/kloppy/.venv/lib/python3.10/site-packages/kloppy-3.7.1-py3.10.egg/kloppy/_providers/statsbomb.py:67: UserWarning: 

You are about to use StatsBomb public data.
By using this data, you are agreeing to the user agreement. 
The user agreement can be found here: https://github.com/statsbomb/open-data/blob/master/LICENSE.pdf

  warnings.warn(

Arturo Vidal is a substitute on the side of Barcelona. We add lineup to all events so we are able to filter out events where Arturo Vidal is on the pitch.

In [38]:

arturo_vidal = home_team.get_player_by_id(8206)

arturo_vidal = home_team.get_player_by_id(8206)

In [39]:

dataframe = (
    dataset
    .add_state('lineup')
    .filter(lambda event: arturo_vidal in event.state['lineup'].players)
    .to_df()
)

dataframe = ( dataset .add_state('lineup') .filter(lambda event: arturo_vidal in event.state['lineup'].players) .to_df() )

In [40]:

print(f"time on pitch: {dataframe['timestamp'].max() - dataframe['timestamp'].min()} seconds")

print(f"time on pitch: {dataframe['timestamp'].max() - dataframe['timestamp'].min()} seconds")

time on pitch: 490.6479999999997 seconds

In [41]:

dataframe = (
    dataset
    .add_state('lineup')
    .filter(lambda event: event.event_type == EventType.PASS and event.team == home_team)
    .to_df(
        "*",
        vidal_on_pitch=lambda event: arturo_vidal in event.state['lineup'].players
    )
)

dataframe = ( dataset .add_state('lineup') .filter(lambda event: event.event_type == EventType.PASS and event.team == home_team) .to_df( "*", vidal_on_pitch=lambda event: arturo_vidal in event.state['lineup'].players ) )

In [42]:

dataframe = dataframe.groupby(['vidal_on_pitch'])['success'].agg(['sum', 'count'])
dataframe['percentage'] = dataframe['sum'] / dataframe['count'] * 100
dataframe

dataframe = dataframe.groupby(['vidal_on_pitch'])['success'].agg(['sum', 'count']) dataframe['percentage'] = dataframe['sum'] / dataframe['count'] * 100 dataframe

Out[42]:

	sum	count	percentage
vidal_on_pitch
False	709	798	88.847118
True	83	88	94.318182

Add state - formation¶

In this example we'll look at adding formation state to all shots.

In [43]:

from kloppy import statsbomb
from kloppy.domain import EventType


dataset = statsbomb.load_open_data()

dataframe = (
    dataset
    .add_state('formation')
    .filter(
        lambda event: event.event_type == EventType.SHOT
    )
    .to_df(
        "*",
        Team=lambda event: str(event.team),
        Formation=lambda event: str(
            event.state['formation'].home 
            if event.team == dataset.metadata.teams[0] 
            else event.state['formation'].away
        )
    )
)

from kloppy import statsbomb from kloppy.domain import EventType dataset = statsbomb.load_open_data() dataframe = ( dataset .add_state('formation') .filter( lambda event: event.event_type == EventType.SHOT ) .to_df( "*", Team=lambda event: str(event.team), Formation=lambda event: str( event.state['formation'].home if event.team == dataset.metadata.teams[0] else event.state['formation'].away ) ) )

/Users/koen/Developer/Projects/PySport/kloppy/.venv/lib/python3.10/site-packages/kloppy-3.7.1-py3.10.egg/kloppy/_providers/statsbomb.py:67: UserWarning: 

You are about to use StatsBomb public data.
By using this data, you are agreeing to the user agreement. 
The user agreement can be found here: https://github.com/statsbomb/open-data/blob/master/LICENSE.pdf

  warnings.warn(

In [44]:

dataframe_stats = (
    dataframe
    .groupby(['Team', 'Formation'])['success']
    .agg(['sum', 'count'])
)
dataframe_stats['Percentage'] = (
    dataframe_stats['sum'] 
    / dataframe_stats['count'] 
    * 100
)
dataframe_stats.rename(
    columns={
        'sum': 'Goals',
        'count': 'Shots'
    }
)

dataframe_stats = ( dataframe .groupby(['Team', 'Formation'])['success'] .agg(['sum', 'count']) ) dataframe_stats['Percentage'] = ( dataframe_stats['sum'] / dataframe_stats['count'] * 100 ) dataframe_stats.rename( columns={ 'sum': 'Goals', 'count': 'Shots' } )

Out[44]:

		Goals	Shots	Percentage
Team	Formation
Barcelona	4-3-3	3	14	21.428571
	4-4-2	0	11	0.000000
Deportivo Alavés	4-1-4-1	0	2	0.000000
	4-4-2	0	1	0.000000