[Football Analysis] 9. 슈팅 데이터 특징 추출

작성일자 : 2023-12-24

Ver 0.1.1

 

In [1]:

# 필요 패키지 불러오기
import os
import pandas as pd
import numpy as np
from tqdm import tqdm

In [2]:

# 현대 디렉토리 확인
current_dir = os.getcwd()
current_dir

Out[2]:

'/Users/limjongjun/Desktop/JayJay/Growth/Python/soccer-analytics/Excercise'

In [3]:

# Data가 있는 디렉토리로 변경
new_dir = '/Users/limjongjun/Desktop/JayJay/Growth/Python/soccer-analytics'
os.chdir(new_dir)
print(current_dir)

/Users/limjongjun/Desktop/JayJay/Growth/Python/soccer-analytics/Excercise

전경기 슈팅 데이터 불러오기¶

(1) 이벤트 데이터 불러오기

• 머신러닝은 데이터가 많이 확보 될 수록 성능에 유리하므로 모든 데이터 불러오기

In [4]:

competitions = [x for x in os.listdir('data/refined_events') if not x.startswith('.')]
match_events_list = []

for competition_name in competitions:
    match_df = pd.read_csv(f'data/refined_events/{competition_name}/matches.csv', index_col=0, encoding='utf-8-sig')

    for match_id in tqdm(match_df.index, desc=f"{competition_name + ' ':10s}"):
        match_events = pd.read_pickle(f'data/refined_events/{competition_name}/{match_id}.pkl')
        match_events['competition_name'] = competition_name # 어떤 대회인지 명시 
        match_events_list.append(match_events)

events = pd.concat(match_events_list, ignore_index=True)
events

England   : 100%|██████████| 380/380 [00:00<00:00, 544.61it/s]
France    : 100%|██████████| 380/380 [00:00<00:00, 546.76it/s]
Spain     : 100%|██████████| 380/380 [00:00<00:00, 640.59it/s]
Germany   : 100%|██████████| 306/306 [00:00<00:00, 529.96it/s]
World_Cup : 100%|██████████| 64/64 [00:00<00:00, 293.18it/s]
Italy     : 100%|██████████| 380/380 [00:00<00:00, 571.97it/s]
European_Championship : 100%|██████████| 51/51 [00:00<00:00, 764.90it/s]

Out[4]:

	match_id	event_id	period	time	team_id	team_name	player_id	player_name	event_type	sub_event_type	tags	start_x	start_y	end_x	end_y	competition_name
0	2499719	177959171	1H	2.759	1609	Arsenal	25413	A. Lacazette	Pass	Simple pass	[Accurate]	50.96	34.68	32.24	14.96	England
1	2499719	177959172	1H	4.947	1609	Arsenal	370224	R. Holding	Pass	High pass	[Accurate]	32.24	14.96	53.04	17.00	England
2	2499719	177959173	1H	6.542	1609	Arsenal	3319	M. Özil	Pass	Head pass	[Accurate]	53.04	17.00	36.40	19.72	England
3	2499719	177959174	1H	8.143	1609	Arsenal	120339	Mohamed Elneny	Pass	Head pass	[Accurate]	36.40	19.72	42.64	3.40	England
4	2499719	177959175	1H	10.302	1609	Arsenal	167145	Bellerín	Pass	Simple pass	[Accurate]	42.64	3.40	74.88	8.16	England
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
3273243	1694440	90589035	E2	984.817	4418	France	7936	P. Pogba	Offside		[]	92.56	48.96	NaN	NaN	European_Championship
3273244	1694440	90589357	E2	1026.418	9905	Portugal	70134	Rui Patrício	Foul	Time lost foul	[Yellow card]	10.40	26.52	NaN	NaN	European_Championship
3273245	1694440	90589358	E2	1031.554	9905	Portugal	70134	Rui Patrício	Free kick	Free kick	[Accurate]	9.36	25.16	69.68	14.96	European_Championship
3273246	1694440	90589359	E2	1035.343	9905	Portugal	70410	Éder	Duel	Air duel	[Lost, Not accurate]	69.68	14.96	69.68	14.96	European_Championship
3273247	1694440	90589040	E2	1036.319	4418	France	7936	P. Pogba	Duel	Air duel	[Won, Accurate]	34.32	53.04	NaN	NaN	European_Championship

3273248 rows × 16 columns

(2) 슈팅 데이터 필터링

In [5]:

shots = events[
    (events['event_type'] == 'Shot') | (events['sub_event_type'].isin(['Free kick shot', 'Penalty']))
].reset_index(drop=True)
shots

Out[5]:

	match_id	event_id	period	time	team_id	team_name	player_id	player_name	event_type	sub_event_type	tags	start_x	start_y	end_x	end_y	competition_name
0	2499719	177959212	1H	94.596	1609	Arsenal	25413	A. Lacazette	Shot	Shot	[Goal, Right foot, Opportunity, Position: Goal...	91.52	40.12	104.00	34.00	England
1	2499719	177959247	1H	179.855	1631	Leicester City	26150	R. Mahrez	Shot	Shot	[Left foot, Opportunity, Position: Out center ...	88.40	32.64	104.00	34.00	England
2	2499719	177959280	1H	254.745	1631	Leicester City	14763	S. Okazaki	Shot	Shot	[Goal, Head/body, Opportunity, Position: Goal ...	99.84	32.64	104.00	34.00	England
3	2499719	177959289	1H	425.824	1609	Arsenal	7868	A. Oxlade-Chamberlain	Shot	Shot	[Left foot, Opportunity, Position: Out high le...	84.24	45.56	104.00	34.00	England
4	2499719	177959429	1H	815.462	1609	Arsenal	7868	A. Oxlade-Chamberlain	Shot	Shot	[Right foot, Opportunity, Position: Goal low l...	78.00	47.60	104.00	34.00	England
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
45940	1694440	90588583	2H	2776.504	4418	France	25575	A. Gignac	Shot	Shot	[Right foot, Opportunity, Position: Post cente...	96.72	39.44	104.00	34.00	European_Championship
45941	1694440	90589205	E1	807.318	9905	Portugal	70410	Éder	Shot	Shot	[Head/body, Opportunity, Position: Goal center...	93.60	42.16	104.00	34.00	European_Championship
45942	1694440	90589242	E2	144.487	9905	Portugal	28907	Raphaël Guerreiro	Free kick	Free kick shot	[Left foot, Direct, Position: Post high left, ...	79.04	46.92	104.00	34.00	European_Championship
45943	1694440	90589254	E2	204.428	9905	Portugal	70410	Éder	Shot	Shot	[Goal, Right foot, Opportunity, Position: Goal...	80.08	36.72	104.00	34.00	European_Championship
45944	1694440	90589034	E2	981.239	4418	France	134513	A. Martial	Shot	Shot	[Right foot, Blocked, Not accurate]	90.48	40.80	94.64	39.44	European_Championship

45945 rows × 16 columns

슈팅별 특징 추출¶

(1) 슈팅 위치 및 거리 계산

In [6]:

shot_features = pd.DataFrame(index=shots.index)
shot_features['x'] = 104 - shots['start_x']

# y축 기준으로 골대는 정중앙에 위치하고 있기 때문에 y의 절반 값인 34을 빼준다. 
shot_features['y'] = shots['start_y'] - 34 

# 피타고라스 정리에 의해 Distance를 구해준다. (np.linalg.norm 함수 적용)
shot_features['distance'] = shot_features[['x', 'y']].apply(np.linalg.norm, axis=1) 
shot_features

Out[6]:

	x	y	distance
0	12.48	6.12	13.899813
1	15.60	-1.36	15.659170
2	4.16	-1.36	4.376665
3	19.76	11.56	22.893038
4	26.00	13.60	29.342120
...	...	...	...
45940	7.28	5.44	9.088014
45941	10.40	8.16	13.219138
45942	24.96	12.92	28.105658
45943	23.92	2.72	24.074152
45944	13.52	6.80	15.133750

45945 rows × 3 columns

(2) 슈팅 각도 계산

• 슈팅 각도도 득점 확률에 기여하는 요소이므로 계산필요

In [7]:

x = shot_features['x']
y = shot_features['y']

goal_width = 7.32

# 각도 theta를 구하는 공식
angles = np.arctan((goal_width * x) / (x ** 2 + y ** 2 - (goal_width / 2) ** 2)) * 180 / np.pi

shot_features['angle'] = np.where(angles >= 0, angles, angles + 180)
shot_features

Out[7]:

	x	y	distance	angle
0	12.48	6.12	13.899813	26.933236
1	15.60	-1.36	15.659170	26.224941
2	4.16	-1.36	4.376665	79.289489
3	19.76	11.56	22.893038	15.813597
4	26.00	13.60	29.342120	12.655803
...	...	...	...	...
45940	7.28	5.44	9.088014	37.600623
45941	10.40	8.16	13.219138	25.258830
45942	24.96	12.92	28.105658	13.240079
45943	23.92	2.72	24.074152	17.184831
45944	13.52	6.80	15.133750	24.652934

45945 rows × 4 columns

(3) 슈팅 유형 및 득점 여부 추출

In [8]:

shot_features['freekick'] = (shots['event_type'] == 'Free kick').astype(int)
shot_features['header'] = shots['tags'].apply(lambda x: 'Head/body' in x).astype(int)
shot_features['goal'] = shots['tags'].apply(lambda x: 'Goal' in x).astype(int)
shot_features

Out[8]:

	x	y	distance	angle	freekick	header	goal
0	12.48	6.12	13.899813	26.933236	0	0	1
1	15.60	-1.36	15.659170	26.224941	0	0	0
2	4.16	-1.36	4.376665	79.289489	0	1	1
3	19.76	11.56	22.893038	15.813597	0	0	0
4	26.00	13.60	29.342120	12.655803	0	0	0
...	...	...	...	...	...	...	...
45940	7.28	5.44	9.088014	37.600623	0	0	0
45941	10.40	8.16	13.219138	25.258830	0	1	0
45942	24.96	12.92	28.105658	13.240079	1	0	0
45943	23.92	2.72	24.074152	17.184831	0	0	1
45944	13.52	6.80	15.133750	24.652934	0	0	0

45945 rows × 7 columns

(4) 슈팅 데이터 연결 및 저장

In [9]:

# 계산한 슈팅 데이터를 앞에서 생성한 DataFrame과 병합 (오른쪽에 붙이는 형식)
shots = pd.concat([shots[['competition_name'] + shots.columns[:-5].tolist()], shot_features], axis=1)
shots

Out[9]:

	competition_name	match_id	event_id	period	time	team_id	team_name	player_id	player_name	event_type	sub_event_type	tags	x	y	distance	angle	freekick	header	goal
0	England	2499719	177959212	1H	94.596	1609	Arsenal	25413	A. Lacazette	Shot	Shot	[Goal, Right foot, Opportunity, Position: Goal...	12.48	6.12	13.899813	26.933236	0	0	1
1	England	2499719	177959247	1H	179.855	1631	Leicester City	26150	R. Mahrez	Shot	Shot	[Left foot, Opportunity, Position: Out center ...	15.60	-1.36	15.659170	26.224941	0	0	0
2	England	2499719	177959280	1H	254.745	1631	Leicester City	14763	S. Okazaki	Shot	Shot	[Goal, Head/body, Opportunity, Position: Goal ...	4.16	-1.36	4.376665	79.289489	0	1	1
3	England	2499719	177959289	1H	425.824	1609	Arsenal	7868	A. Oxlade-Chamberlain	Shot	Shot	[Left foot, Opportunity, Position: Out high le...	19.76	11.56	22.893038	15.813597	0	0	0
4	England	2499719	177959429	1H	815.462	1609	Arsenal	7868	A. Oxlade-Chamberlain	Shot	Shot	[Right foot, Opportunity, Position: Goal low l...	26.00	13.60	29.342120	12.655803	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
45940	European_Championship	1694440	90588583	2H	2776.504	4418	France	25575	A. Gignac	Shot	Shot	[Right foot, Opportunity, Position: Post cente...	7.28	5.44	9.088014	37.600623	0	0	0
45941	European_Championship	1694440	90589205	E1	807.318	9905	Portugal	70410	Éder	Shot	Shot	[Head/body, Opportunity, Position: Goal center...	10.40	8.16	13.219138	25.258830	0	1	0
45942	European_Championship	1694440	90589242	E2	144.487	9905	Portugal	28907	Raphaël Guerreiro	Free kick	Free kick shot	[Left foot, Direct, Position: Post high left, ...	24.96	12.92	28.105658	13.240079	1	0	0
45943	European_Championship	1694440	90589254	E2	204.428	9905	Portugal	70410	Éder	Shot	Shot	[Goal, Right foot, Opportunity, Position: Goal...	23.92	2.72	24.074152	17.184831	0	0	1
45944	European_Championship	1694440	90589034	E2	981.239	4418	France	134513	A. Martial	Shot	Shot	[Right foot, Blocked, Not accurate]	13.52	6.80	15.133750	24.652934	0	0	0

45945 rows × 19 columns

In [10]:

shots.to_pickle('data/shots.pkl')