#!/usr/bin/env python
# coding: utf-8

# # Which Teams are Failing to Meet Expectations?

# In[135]:


import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
import matplotlib.patheffects as path_effects
import matplotlib.font_manager as fm
import matplotlib.colors as mcolors
from matplotlib import cm
from highlight_text import fig_text, ax_text
from matplotlib.colors import LinearSegmentedColormap, Normalize
from matplotlib import cm
import matplotlib.gridspec as gridspec
import numpy as np

from PIL import Image
import urllib
import json
import os


# In[136]:


font_path = "../assets/fonts"
for x in os.listdir(font_path):
    for y in os.listdir(f"{font_path}/{x}"):
        if y.split(".")[-1] == "ttf":
            fm.fontManager.addfont(f"{font_path}/{x}/{y}")
            try:
                fm.FontProperties(weight=y.split("-")[-1].split(".")[0].lower(), fname=y.split("-")[0])
            except Exception:
                continue

plt.style.use("../assets/stylesheets/soc_base.mplstyle")


# ## The data
# 
# Here we read 538's predictions and compute the expected point tally of each side, using the pre-match probabilities assigned by 538.

# In[137]:


df = pd.read_csv('data/10032022.csv', index_col=0).reset_index(drop=True)
df


# Custom function to compute points across teams.

# In[138]:


def compute_points(x,y):
    '''
    This function computed the points achieved by each side.
    '''
    if x > y:
        return 3
    if x == y:
        return 1
    else:
        return 0


# In[139]:


df['xpoints1'] = df['prob1']*3 + df['probtie']
df['xpoints2'] = df['prob2']*3 + df['probtie']
df['points1'] = [compute_points(x,y) for x,y in zip(df['score1'], df['score2'])]
df['points2'] = [compute_points(y,x) for x,y in zip(df['score1'], df['score2'])]
df


# In[140]:


df_team1 = df[['team1', 'xpoints1', 'points1']]
df_team2 = df[['team2', 'xpoints2', 'points2']]
df_team1.columns = ['team', 'xpoints', 'points']
df_team2.columns = ['team', 'xpoints', 'points']
df_g = (
    pd.concat([df_team1, df_team2])
    .groupby(['team']).sum()
    .reset_index()
    .assign(difference = lambda x: x.points - x.xpoints)
)
df_g.sort_values(by='difference')


# Now we get the twelve teams that have most underperformed their expected points.

# In[141]:


df_top_10 = df_g.sort_values(by='difference').head(12).reset_index(drop=True)
df_top_10


# For the logos, we need to do some manual labour and get the team's Fotmob ID.

# In[142]:


team_map = {
    'team': [
        'Strasbourg',
        'Bayer Leverkusen',
        'Leicester City',
        'VfL Bochum',
        'Sampdoria',
        'AC Ajaccio',
        'Liverpool',
        'Bayern Munich',
        'Internazionale',
        'Elche',
        'VfB Stuttgart',
        'Wolverhampton'
    ],
    'fotmob_id':[
        9848,
        8178,
        8197,
        9911,
        9882,
        8576,
        8650,
        9823,
        8636,
        10268,
        10269,
        8602
    ],
    'country':[
        'fra',
        'deu',
        'eng',
        'deu',
        'ita',
        'fra',
        'eng',
        'deu',
        'ita',
        'esp',
        'deu',
        'eng'
    ]
}

team_map = pd.DataFrame(team_map)
team_map


# In[143]:


df_top_10 = pd.merge(df_top_10, team_map)
df_top_10


# Finally, we create a function that takes the timeline of this xPoints and actual points so we can plot a time series.

# In[144]:


def get_cumpoints_df(team, data = df):

    df = data.copy()
    df['date'] = pd.to_datetime(df['date'], format='%Y-%m-%d')
    is_in_home = df[(df['team1'] == team)][['team1', 'points1', 'xpoints1', 'date']].reset_index(drop=True)
    is_in_away = df[(df['team2'] == team)][['team2', 'points2', 'xpoints2', 'date']].reset_index(drop=True)
    is_in_home.columns = ['team', 'points', 'xpoints', 'date']
    is_in_away.columns = ['team', 'points', 'xpoints', 'date']
    df = pd.concat([is_in_home, is_in_away]).sort_values(by='date', ascending=True).reset_index(drop=True)
    df['cum_points'] = df['points'].cumsum()
    df['cum_xpoints'] = df['xpoints'].cumsum()
    return df

# Check to see if it works
get_cumpoints_df('Strasbourg')


# ## The Visual

# In[145]:


def plot_point_difference(ax, team, label_y = False, data=df):
    ax.grid(ls='--', color='lightgrey')
    for spine in ax.spines.values():
        spine.set_edgecolor('lightgrey')
        spine.set_linestyle('dashed')
    ax.tick_params(color='lightgrey', labelsize=8, labelcolor='grey')
    test_df = get_cumpoints_df(team, data)

    ax.plot(test_df.index + 1, test_df['cum_points'], marker='o', mfc='white', ms=5, color='#287271')
    ax.plot(test_df.index + 1, test_df['cum_xpoints'], marker='o', mfc='white', ms=5, color='#D81159')
    ax.yaxis.set_major_locator(ticker.MultipleLocator(3))
    ax.yaxis.set_major_formatter(ticker.StrMethodFormatter('{x:.0f}'))
    ax.set_ylim(-1,22)
    ax.set_xlim(0.25)
    ax.set_xticklabels([])
    mid_point = test_df['cum_points'].iloc[-1] + (test_df['cum_xpoints'].iloc[-1] - test_df['cum_points'].iloc[-1])/2
    difference = test_df['cum_points'].iloc[-1] - test_df['cum_xpoints'].iloc[-1]
    text_ = ax.annotate(
        xy=(ax.get_xlim()[1], mid_point),
        text=f'{difference:.1f}',
        xytext=(-5,0),
        ha='center',
        va='center',
        color='#D81159',
        weight='bold',
        size=7,
        textcoords='offset points'
    )
    text_.set_path_effects(
        [path_effects.Stroke(linewidth=1.5, foreground='white'), path_effects.Normal()]
    )
    # ----------------------------------------------------------------
    if label_y:
        ax.set_ylabel('Points', color='grey', size=8)
    else:
        ax.set_yticklabels([])


# In[146]:


fig = plt.figure(figsize=(9, 7), dpi = 200)
nrows = 6
ncols = 4
gspec = gridspec.GridSpec(
    ncols=ncols, nrows=nrows, figure=fig, 
    height_ratios = [(1/nrows)*2.6 if x % 2 != 0 else (1/nrows)/2.6 for x in range(nrows)], hspace=0.2
)

plot_counter = 0
logo_counter = 0
for row in range(nrows):
    for col in range(ncols):
        if row % 2 != 0:
            ax = plt.subplot(
                gspec[row, col],
                facecolor = "#EFE9E6"
            )
            team = df_top_10['team'].iloc[plot_counter]
            if col == 0:
                label_y = True
            else:
                label_y = False
            
            plot_point_difference(ax, team, label_y)           
            plot_counter += 1
        else:
            teamId = df_top_10['fotmob_id'].iloc[logo_counter]
            teamName = df_top_10['team'].iloc[logo_counter]
            xpoints = df_top_10['xpoints'].iloc[logo_counter]
            apoints = df_top_10['points'].iloc[logo_counter]
            fotmob_url = 'https://images.fotmob.com/image_resources/logo/teamlogo/'
            logo_ax = plt.subplot(
                gspec[row,col],
                anchor = 'NW', facecolor = '#EFE9E6'
            )
            club_icon = Image.open(urllib.request.urlopen(f'{fotmob_url}{teamId:.0f}.png')).convert('LA')
            logo_ax.imshow(club_icon)
            logo_ax.axis('off')
            # -- Add the team name
            ax_text(
                x = 1.2, 
                y = 0.7,
                s = f'<{teamName}>\n<Exp. points: {xpoints:.1f}> <|> <Act. points: {apoints:.0f}>',
                ax = logo_ax, 
                highlight_textprops=[{'weight':'bold', 'font':'DM Sans'},{'size':'7', 'color': '#D81159'}, {'size':'7', 'color': 'grey'}, {'size':'7', 'color': '#287271'}],
                font = 'Karla', 
                ha = 'left', 
                size = 9, 
                annotationbbox_kw = {'xycoords':'axes fraction'}
            )
            logo_counter += 1

fig_text(
    x=0.12, y=.95, 
    s='Which Teams are Failing to Meet Expectations?',
    va='bottom', ha='left',
    fontsize=15, color='black', font='DM Sans', weight='bold'
)
fig_text(
	x=0.12, y=.92, 
    s='<Expected points> (in probabilistic terms) vs. <actual points> | Probabilities taken from FiveThirtyEight | viz by @sonofacorner',
    highlight_textprops=[{'weight':'bold', 'color': '#D81159'}, {'weight':'bold', 'color': '#287271'}],
	va='bottom', ha='left',
	fontsize=8, color='#4E616C', font='Karla'
)

plt.savefig(
	"figures/10032022_underperformers.png",
	dpi = 600,
	facecolor = "#EFE9E6",
	bbox_inches="tight",
    edgecolor="none",
	transparent = False
)

plt.savefig(
	"figures/10032022_underperformers_tr.png",
	dpi = 600,
	facecolor = "none",
	bbox_inches="tight",
    edgecolor="none",
	transparent = True
)