Fantasy Football
ESPN Fantasy Football from the league I’ve played in 2023.
Data extracted using espnfantasyfootball.
Initiate
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
df = pd.read_csv("data/players.csv")
mt = pd.read_csv("data/matches.csv")
Some Variables to work with
my_team = "Steely Dan Fan Club"
last_full_week = 17
Prepare
- Rename columns
- Filter only played weeks
# 1
df.rename(columns={
"Week": "week",
"PlayerName": "name",
"PlayerScoreActual": "score",
"PlayerScoreProjected": "projected",
"PlayerFantasyTeam": "team_index",
"PlayerRosterSlot": "position",
"TeamName": "team",
"FullName": "user"
}, inplace=True)
mt.rename(columns={
"Week": "week",
"Name1": "team1",
"Score1": "score1",
"Name2": "team2",
"Score2": "score2",
"Type": "type"
}, inplace=True)
#2
df = df[df.week<=last_full_week]
mt = mt[mt.week<=last_full_week].drop_duplicates()
Working Data
df.info()
df.head()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3828 entries, 0 to 3827
Data columns (total 8 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 week 3828 non-null int64
1 name 3828 non-null object
2 score 3828 non-null float64
3 projected 3828 non-null float64
4 team_index 3828 non-null int64
5 position 3828 non-null object
6 team 3828 non-null object
7 user 3828 non-null object
dtypes: float64(2), int64(2), object(4)
memory usage: 239.4+ KB
| week | name | score | projected | team_index | position | team | user | |
|---|---|---|---|---|---|---|---|---|
| 0 | 1 | Davante Adams | 15.60 | 21.204351 | 1 | WR | Joe Burrow's Side Piece | Briony Quirk |
| 1 | 1 | Patrick Mahomes | 18.54 | 24.679680 | 1 | QB | Joe Burrow's Side Piece | Briony Quirk |
| 2 | 1 | Aaron Jones | 28.70 | 16.338186 | 1 | RB | Joe Burrow's Side Piece | Briony Quirk |
| 3 | 1 | T.J. Hockenson | 12.00 | 13.890898 | 1 | TE | Joe Burrow's Side Piece | Briony Quirk |
| 4 | 1 | Brandon Aiyuk | 37.90 | 11.752372 | 1 | WR | Joe Burrow's Side Piece | Briony Quirk |
mt.info() # the 2 null team1 were a Bye week in the playoffs
mt.head()
<class 'pandas.core.frame.DataFrame'>
Index: 120 entries, 0 to 119
Data columns (total 6 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 week 120 non-null int64
1 team1 118 non-null object
2 score1 118 non-null float64
3 team2 120 non-null object
4 score2 120 non-null float64
5 type 120 non-null object
dtypes: float64(2), int64(1), object(3)
memory usage: 6.6+ KB
| week | team1 | score1 | team2 | score2 | type | |
|---|---|---|---|---|---|---|
| 0 | 1 | Luluists | 141.64 | Briony’s Blownbackers | 160.76 | Regular |
| 1 | 1 | Steely Dan Fan Club | 132.24 | Team Sand | 61.10 | Regular |
| 2 | 1 | Intentional Sounding | 106.98 | Joey's Calzones | 77.76 | Regular |
| 3 | 1 | RB Strike Supporter | 89.34 | Joe Burrow's Side Piece | 133.44 | Regular |
| 4 | 1 | THE FLOCK | 112.28 | Crimson N Cream Deng Xiaopings | 120.94 | Regular |
Metrics, Tables and Charts
Teams wins and total scores
wins = pd.Series(np.where(mt.score1 > mt.score2, mt.team1, mt.team2)).value_counts()
scores = df[df.position!='Bench'].groupby(['team']).score.sum().sort_values(ascending=False)
pd.concat({'wins': wins, 'total_scores': scores}, axis=1).sort_values(['wins', 'total_scores'], ascending=False)
# To get the average point per match (week) per team
# df[df.position!='Bench'].groupby(['week', 'team']).score.sum().groupby('team').mean().sort_values(ascending=False)
| wins | total_scores | |
|---|---|---|
| SlurpmystepBURROW | 14 | 2244.36 |
| DEVOLDER DOMINATORS | 11 | 2099.04 |
| Joe Burrow's Side Piece | 10 | 2192.42 |
| Honolulu HSTSchads | 10 | 1994.64 |
| Briony’s Blownbackers | 10 | 1828.12 |
| Team Sand | 9 | 2136.64 |
| RB Strike Supporter | 8 | 2079.32 |
| Steely Dan Fan Club | 8 | 2026.50 |
| Crimson N Cream Deng Xiaopings | 8 | 2011.84 |
| Intentional Sounding | 7 | 2093.34 |
| Luluists | 7 | 1972.04 |
| Joey's Calzones | 7 | 1909.52 |
| non-human biologics | 6 | 1832.32 |
| THE FLOCK | 5 | 1700.28 |
132.02117647058824
Teams’ average of points per game
df[df.position!='Bench'].groupby(['week', 'team']).score.sum().groupby('team').mean().sort_values(ascending=False)
team
SlurpmystepBURROW 132.021176
Joe Burrow's Side Piece 128.965882
Team Sand 125.684706
DEVOLDER DOMINATORS 123.472941
Intentional Sounding 123.137647
RB Strike Supporter 122.312941
Steely Dan Fan Club 119.205882
Crimson N Cream Deng Xiaopings 118.343529
Honolulu HSTSchads 117.331765
Luluists 116.002353
Joey's Calzones 112.324706
non-human biologics 107.783529
Briony’s Blownbackers 107.536471
THE FLOCK 100.016471
Name: score, dtype: float64
Or, checking individually
team_name = "SlurpmystepBURROW"
pd.concat([mt.loc[mt['team1']==team_name, 'score1'], mt.loc[mt['team2']==team_name, 'score2']]).mean()
132.02117647058824
Comparing players
def plot_players_score_throughout_season(df, ax, players, title="Players performances throughout the season"):
""" Given the DataFrame, the axes, and the players names, plot the scores of the players throughout the weeks of the season in the axes
"""
df[df.name.isin(players)][['week', 'name', 'score']]\
.pivot(columns="name", index="week", values="score")\
.plot(ax=ax)
ax.set_title(title)
ax.legend(title=None)
# ax.set_ylim(bottom=0)
# Set xticks to be only integers
from matplotlib.ticker import MaxNLocator
ax.xaxis.set_major_locator(MaxNLocator(integer=True))
%matplotlib notebook
rbs = ["Bijan Robinson", "Miles Sanders"]
wrs = ["Amon-Ra St. Brown", "Chris Olave", "Amari Cooper", "Amari Cooper", "Jakobi Meyers", "DeVante Parker", "Zay Jones"]
qbs = ["Tua Tagovailoa", "Jordan Love"]
fig, ax = plt.subplots(figsize=(16,4), ncols=3)
plot_players_score_throughout_season(df[(df.team==my_team)], ax[0], rbs, title="Running Back performances")
plot_players_score_throughout_season(df[(df.team==my_team)], ax[1], qbs, title="Quarterbacks performances")
plot_players_score_throughout_season(df[(df.team==my_team)], ax[2], wrs, title="Wide Receivers performances")
ax[2].legend(loc='lower center', bbox_to_anchor=(1.3, 0.2))
<IPython.core.display.Javascript object>
<matplotlib.legend.Legend at 0x7fb5e72843d0>
%matplotlib inline
Comparing projections with actual scores
import matplotlib.ticker as ticker
fig, axs = plt.subplots(figsize=(14, 6), ncols=2)
df['projection_error'] = df.projected - df.score
sns.scatterplot(data=df[df.position!='Bench'], x='score', y='projected', hue='position', ax=axs[0])
axs[0].set(title="Scatter of projected score VS actual score", xlabel="Actual Score", ylabel="Projected Score")
# Y axis limits to the same as X axis limits
axs[0].set_ylim(bottom=axs[0].get_xlim()[0], top=axs[0].get_xlim()[1])
df['projection_error'].plot.hist(bins=30, ax=axs[1])
axs[1].xaxis.set_major_locator(ticker.MultipleLocator(5))
axs[1].set(title="Histogram of projection error (projected - actual score)", xlabel="Error")
[Text(0.5, 1.0, 'Histogram of projection error (projected - actual score)'),
Text(0.5, 0, 'Error')]
Comparing a positions scores across teams in a week
week = 12
position = 'QB'
rb_score_week = df[(df.week==week) & (df.position==position)][["team", "name", "score"]]\
.sort_values(by="team")\
.set_index(["team", "name"])
fig, ax = plt.subplots(figsize=(12, 4))
rb_score_week.plot.bar(ax=ax, legend=False)
ax.set_title(f"{position} scores in week {week}")
#ax.set_xticks(ax.get_xticks(), labels=rb_score_week.index.get_level_values(1))
ax.set_xlabel("Players")
ax.set_ylabel("Points")
plt.show()
