Stats Project - Soccer¶

by Flor Nightgale¶

For this project we used secondary data about Premier League (Soccer).

Team Statistics¶

import pandas as pd
data = pd.read_html('https://www.espn.com/soccer/table/_/league/eng.1')
teams = data[0].join(data[1]) # join the two data tables together
teams

	2020-2021	GP	W	L	F	A	GD	P
0	1LEILeicester City	2	2	0	7	2	5	6
1	2EVEEverton	2	2	0	6	2	4	6
2	3ARSArsenal	2	2	0	5	1	4	6
3	4LIVLiverpool	2	2	0	6	3	3	6
4	5CRYCrystal Palace	2	2	0	4	1	3	6
5	6TOTTottenham Hotspur	2	1	1	5	3	2	3
6	7MNCManchester City	1	1	0	3	1	2	3
7	8BHABrighton & Hove Albion	2	1	1	4	3	1	3
8	9AVLAston Villa	1	1	0	1	0	1	3
9	10LEELeeds United	2	1	1	7	7	0	3
10	11CHEChelsea	2	1	1	3	3	0	3
11	12WOLWolverhampton Wanderers	2	1	1	3	3	0	3
12	13NEWNewcastle United	2	1	1	2	3	-1	3
13	14BURBurnley	1	0	1	2	4	-2	0
14	15MANManchester United	1	0	1	1	3	-2	0
15	16WHUWest Ham United	2	0	2	1	4	-3	0
16	17SHUSheffield United	2	0	2	0	3	-3	0
17	18FULFulham	2	0	2	3	7	-4	0
18	19SOUTSouthampton	2	0	2	2	6	-4	0
19	20WBAWest Bromwich Albion	2	0	2	2	8	-6	0

Columns in the data set are:

GP: Games Played
W: Wins
D: Draws
L: Losses
F: Goals For
A: Goals Against
GD: Goal Difference
P: Points

Notice that the ranking (index values) start at zero. As well, the team names got combined with their ranks and abbreviations, let’s cut those out and leave just the team names.

For each team name, the second character is a lowercase letter, so we’ll find the first lowercase letter then take just the characters from one before that until the end of the name.

We’ll also rename the columns.

for i, row in teams.iterrows():
    for character in row[0]:
        if character.islower(): # we've found the first lowercase letter
            start_here = row[0].index(character)-1
            team_name = row[0][start_here:]
            break # stop looking through the team name
    teams.iloc[i,0] = team_name
teams.columns = ['Team','Games Played','Wins','Draws','Losses','Goals For','Goals Against','Goal Difference','Points']
teams

	Team	Games Played	Wins	Losses	Goals For	Goals Against	Goal Difference	Points
0	Leicester City	2	2	0	7	2	5	6
1	Everton	2	2	0	6	2	4	6
2	Arsenal	2	2	0	5	1	4	6
3	Liverpool	2	2	0	6	3	3	6
4	Crystal Palace	2	2	0	4	1	3	6
5	Tottenham Hotspur	2	1	1	5	3	2	3
6	Manchester City	1	1	0	3	1	2	3
7	Brighton & Hove Albion	2	1	1	4	3	1	3
8	Aston Villa	1	1	0	1	0	1	3
9	Leeds United	2	1	1	7	7	0	3
10	Chelsea	2	1	1	3	3	0	3
11	Wolverhampton Wanderers	2	1	1	3	3	0	3
12	Newcastle United	2	1	1	2	3	-1	3
13	Burnley	1	0	1	2	4	-2	0
14	Manchester United	1	0	1	1	3	-2	0
15	West Ham United	2	0	2	1	4	-3	0
16	Sheffield United	2	0	2	0	3	-3	0
17	Fulham	2	0	2	3	7	-4	0
18	Southampton	2	0	2	2	6	-4	0
19	West Bromwich Albion	2	0	2	2	8	-6	0

Statistical Calculations¶

The describe() method does some statisical calculations for us.

team_stats = teams.describe()
team_stats

	Games Played	Wins	Draws	Losses	Goals For	Goals Against	Goal Difference	Points
count	20.000000	20.000000	20.0	20.000000	20.000000	20.000000	20.000000	20.000000
mean	1.800000	0.900000	0.0	0.900000	3.350000	3.350000	0.000000	2.700000
std	0.410391	0.788069	0.0	0.788069	2.084403	2.158825	3.077935	2.364207
min	1.000000	0.000000	0.0	0.000000	0.000000	0.000000	-6.000000	0.000000
25%	2.000000	0.000000	0.0	0.000000	2.000000	2.000000	-2.250000	0.000000
50%	2.000000	1.000000	0.0	1.000000	3.000000	3.000000	0.000000	3.000000
75%	2.000000	1.250000	0.0	1.250000	5.000000	4.000000	2.250000	3.750000
max	2.000000	2.000000	0.0	2.000000	7.000000	8.000000	5.000000	6.000000

We can also find the median values.

teams.median()

Games Played       2.0
Wins               1.0
Draws              0.0
Losses             1.0
Goals For          3.0
Goals Against      3.0
Goal Difference    0.0
Points             3.0
dtype: float64

The Goal Difference column is probably the most interesting, since it has the largest range and highest standard deviation. The top teams scored a lot more than they were scored on, and the bottom teams were scored on a lot more than they scored.

Since we are looking at data for all of the teams, we see that the mean number of wins is equal to the mean number of losses. The same goes for goals scored and goals scored against.

import plotly_express as px
fig = px.bar(team_stats.iloc[3:], y='Goal Difference', title='')
fig.show()

If we want to see which teams scored more than the mean value of “Goals For”, we can use the following code.

gf_mean = teams['Goals For'].mean()
teams[teams['Goals For'] > gf_mean]

	Team	Games Played	Wins	Losses	Goals For	Goals Against	Goal Difference	Points
0	Leicester City	2	2	0	7	2	5	6
1	Everton	2	2	0	6	2	4	6
2	Arsenal	2	2	0	5	1	4	6
3	Liverpool	2	2	0	6	3	3	6
4	Crystal Palace	2	2	0	4	1	3	6
5	Tottenham Hotspur	2	1	1	5	3	2	3
7	Brighton & Hove Albion	2	1	1	4	3	1	3
9	Leeds United	2	1	1	7	7	0	3

In general, but not always, the top teams scored more than the average number of goals.

Mean is probably the best measure of central tendency here, since using the median would just give us the top half of the teams. Mode wouldn’t be useful because there aren’t a lot of repeated values in the column.

Let’s see if the top teams had fewer than the mean number of goals scored against them.

ga_mean = teams['Goals Against'].mean()
teams[teams['Goals Against'] < gf_mean]

	Team	Games Played	Wins	Losses	Goals For	Goals Against	Goal Difference	Points
0	Leicester City	2	2	0	7	2	5	6
1	Everton	2	2	0	6	2	4	6
2	Arsenal	2	2	0	5	1	4	6
3	Liverpool	2	2	0	6	3	3	6
4	Crystal Palace	2	2	0	4	1	3	6
5	Tottenham Hotspur	2	1	1	5	3	2	3
6	Manchester City	1	1	0	3	1	2	3
7	Brighton & Hove Albion	2	1	1	4	3	1	3
8	Aston Villa	1	1	0	1	0	1	3
10	Chelsea	2	1	1	3	3	0	3
11	Wolverhampton Wanderers	2	1	1	3	3	0	3
12	Newcastle United	2	1	1	2	3	-1	3
14	Manchester United	1	0	1	1	3	-2	0
16	Sheffield United	2	0	2	0	3	-3	0

Again, it is generally true that the top teams that had fewer goals scored against them.

Teams Visualizations¶

Let’s create some plots of Wins, Losses, Draws versus team rank.

columns = ['Wins', 'Losses', 'Draws']
for column in columns:
    fig = px.scatter(teams, x=teams.index, y=column, title=column+' vs Rank', hover_data=['Team'])
    fig.show()

Player Data¶

We are also going to look at individual player data for scoring and assists. We’ll download both and then look first at the top 10, head(10), of the scorers data table.

stats = pd.read_html('https://www.espn.com/soccer/stats/_/league/ENG.1/view/scoring')
scorers = stats[0]
assists = stats[1]
scorers.head(10)

	RK	Name	Team	P	G
0	1.0	Dominic Calvert-Lewin	Everton	2	4
1	NaN	Son Heung-Min	Tottenham Hotspur	2	4
2	3.0	Mohamed Salah	Liverpool	2	3
3	NaN	Wilfried Zaha	Crystal Palace	2	3
4	5.0	Hélder Costa	Leeds United	2	2
5	NaN	Aleksandar Mitrovic	Fulham	2	2
6	NaN	Neal Maupay	Brighton & Hove Albion	2	2
7	NaN	Sadio Mané	Liverpool	2	2
8	NaN	Patrick Bamford	Leeds United	2	2
9	NaN	Raúl Jiménez	Wolverhampton Wanderers	2	2

Columns:

RK: Ranking
P: Games played
G: Goals scored
A: Assists

There are quite a few missing (NaN) values, which means that player is tied with the player above them, so we can use fillna(method='ffill') which means “forward fill” values to replace missing values.

scorers = scorers.fillna(method='ffill')
scorers.head(10)

	RK	Name	Team	P	G
0	1.0	Dominic Calvert-Lewin	Everton	2	4
1	1.0	Son Heung-Min	Tottenham Hotspur	2	4
2	3.0	Mohamed Salah	Liverpool	2	3
3	3.0	Wilfried Zaha	Crystal Palace	2	3
4	5.0	Hélder Costa	Leeds United	2	2
5	5.0	Aleksandar Mitrovic	Fulham	2	2
6	5.0	Neal Maupay	Brighton & Hove Albion	2	2
7	5.0	Sadio Mané	Liverpool	2	2
8	5.0	Patrick Bamford	Leeds United	2	2
9	5.0	Raúl Jiménez	Wolverhampton Wanderers	2	2

assists = assists.fillna(method='ffill')
assists.head()

	RK	Name	Team	P	A
0	1.0	Harry Kane	Tottenham Hotspur	2	4
1	2.0	Daniel Podence	Wolverhampton Wanderers	2	2
2	2.0	Willian	Arsenal	2	2
3	2.0	Richarlison	Everton	2	2
4	5.0	Tariq Lamptey	Brighton & Hove Albion	2	1

Let’s create histograms for these two data sets.

fig1 = px.histogram(scorers, x='G', title='Histogram of Goals Scored by Top Players')
fig1.show()
fig2 = px.histogram(assists, x='A', title='Histogram of Assists by Top Players')
fig2.show()

Both of these histograms show that there are many more players that scored (or assisted) fewer goals, so the data are not normally distributed.

Research Question¶

Does having more top scoring or top assisting players on a team mean that team has a higher standing?

To answer this question, we will need to group the player data by team and merge the two data tables together. We’ll also drop the columns that we don’t need.

# group the data by team
scorers_team = scorers.groupby('Team').count().drop(columns=['RK', 'Name', 'P'])
assists_team = assists.groupby('Team').count().drop(columns=['RK', 'Name', 'P'])
# merge the players data tables
players = scorers_team.merge(assists_team, on='Team')
# create a column that adds goals and assists
players['Goals and Assists'] = players['G']+players['A']
# sort the values, create an index column, and display the data
players = players.sort_values('Goals and Assists', ascending=False).reset_index()
players

	Team	G	A	Goals and Assists
0	Leeds United	5	5	10
1	Brighton & Hove Albion	4	5	9
2	Arsenal	4	4	8
3	Leicester City	5	3	8
4	Crystal Palace	3	4	7
5	Everton	3	4	7
6	Chelsea	3	2	5
7	Liverpool	3	2	5
8	Manchester City	3	2	5
9	Newcastle United	2	3	5
10	Southampton	2	3	5
11	West Bromwich Albion	2	3	5
12	Fulham	2	2	4
13	Tottenham Hotspur	2	2	4
14	Wolverhampton Wanderers	2	2	4
15	Burnley	2	1	3
16	West Ham United	1	2	3
17	Aston Villa	1	1	2

Now we need to merge this data table with the Teams data table from earlier.

combined_data = teams.merge(players, on='Team', how='left') # left means keep the order from the teams data table
combined_data

	Team	Games Played	Wins	Losses	Goals For	Goals Against	Goal Difference	Points	G	A	Goals and Assists
0	Leicester City	2	2	0	7	2	5	6	5.0	3.0	8.0
1	Everton	2	2	0	6	2	4	6	3.0	4.0	7.0
2	Arsenal	2	2	0	5	1	4	6	4.0	4.0	8.0
3	Liverpool	2	2	0	6	3	3	6	3.0	2.0	5.0
4	Crystal Palace	2	2	0	4	1	3	6	3.0	4.0	7.0
5	Tottenham Hotspur	2	1	1	5	3	2	3	2.0	2.0	4.0
6	Manchester City	1	1	0	3	1	2	3	3.0	2.0	5.0
7	Brighton & Hove Albion	2	1	1	4	3	1	3	4.0	5.0	9.0
8	Aston Villa	1	1	0	1	0	1	3	1.0	1.0	2.0
9	Leeds United	2	1	1	7	7	0	3	5.0	5.0	10.0
10	Chelsea	2	1	1	3	3	0	3	3.0	2.0	5.0
11	Wolverhampton Wanderers	2	1	1	3	3	0	3	2.0	2.0	4.0
12	Newcastle United	2	1	1	2	3	-1	3	2.0	3.0	5.0
13	Burnley	1	0	1	2	4	-2	0	2.0	1.0	3.0
14	Manchester United	1	0	1	1	3	-2	0	NaN	NaN	NaN
15	West Ham United	2	0	2	1	4	-3	0	1.0	2.0	3.0
16	Sheffield United	2	0	2	0	3	-3	0	NaN	NaN	NaN
17	Fulham	2	0	2	3	7	-4	0	2.0	2.0	4.0
18	Southampton	2	0	2	2	6	-4	0	2.0	3.0	5.0
19	West Bromwich Albion	2	0	2	2	8	-6	0	2.0	3.0	5.0

To see if there is a relationship between Goals and Assists and team rank, let’s create another scatterplot.

fig = px.scatter(combined_data, y='Goals and Assists', x=combined_data.index, hover_data=['Team'], title='Goals and Assists vs Team Rank')
fig.show()

Conclusion¶

It looks like higher ranked teams (lower \(x\) values) tend to have more players with more goals and assists, although there is a fair amount of variation in the data.

Perhaps we could look at a similar analysis using a larger data set from a league such as the National Hockey League where there are more games played by more teams.

Callysto Curriculum Notebooks