dataviz.qmd

---
title: "Data Visualization"
format: html
---

```{python}
#| echo: false
#| warning: false
#| message: false

# Imports
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
import matplotlib.patches as mpatches
from scipy import stats

# Loading in data
data = pd.read_csv('./data/cleaned_data.csv')

# Converting date to datetime
data["Date"] = pd.to_datetime(data["Date"])
```

## Making Visualization 1

### Getting Rolling Averages for Visualization 1
```{python}
# This code chunk retrieves the n-game (7-game) rolling average of proportion of low, med, and high band jumps

# Empty list to hold rolling_averages data
rolling_averages = []

# Variable for number of games in rolling average
window = 7

# Looping through each group of the df, where the groups are by player
for player, group in data.groupby("About"):
    # Ensuring the dates are in the correct order for the rolling average calculation
    group = group.sort_values(by = "Date")

    # Calculating the rolling averages, over window games, and assigning that to the new column made for that value
    group["Prop_Jump_High"]= group["Prop_Jump_High"].rolling(window, min_periods = 1).mean()
    group["Prop_Jump_Med"]= group["Prop_Jump_Med"].rolling(window, min_periods = 1).mean()
    group["Prop_Jump_Low"]= group["Prop_Jump_Low"].rolling(window, min_periods = 1).mean()

    # Appending the group to the list of groups to be put in a df later
    rolling_averages.append(group)

# Concatenating the rolling average groups. Each player/group is essentially a row in this new dataframe
rolling_df = pd.concat(rolling_averages)

# Filtering out players who have played less than 15 games, as our analysis requires a pretty large sample size

# Getting game count by player
player_games = rolling_df["About"].value_counts()
# Getting the players who have played more than 15 games
players_met_min_games = player_games[player_games >= 15].index
# Limiting rolling_df to only players in the list of players who have played 15 games
rolling_df = rolling_df[rolling_df["About"].isin(players_met_min_games)]
```

### Creating Visualization 1 Color Palette
```{python}
# Getting the number of seasons per player, and getting a list of players who are in both seasons
players_seasons = rolling_df.groupby('About')['Season'].nunique()
players_in_both = players_seasons[players_seasons > 1].index.tolist()

# Getting list of all players
players = rolling_df['About'].unique()

# Making a palette where players in 1 season are orange, and players in 2 seasons are navy
palette = {player : ('navy' if player in players_in_both else 'orange') for player in players}
```

### Generating and Saving Visualization 1

```{python}
# Creating Figure and Flattening Axes
fig, axs = plt.subplots(1, 2, figsize = (30, 10), sharey = True)
axs = axs.flatten()

# Looping through subplots and seasons
for ax, season in zip(axs, [1, 2]):
    # Filter data for the current season
    season_data = rolling_df[rolling_df["Season"] == season]
    
    # Calculate dates for x-ticks
    date_percentiles = np.percentile(season_data["Date"].values.astype('int64'), [0, 25, 50, 75, 100])
    date_ticks = pd.to_datetime(date_percentiles)

    # Plotting
    plot = sns.lineplot(data=season_data, x='Date', y="Prop_Jump_High", hue="About", palette=palette, ax=ax)
    plot_c = sns.lineplot(data=season_data[season_data["About"] == "Athlete C"], x="Date", y="Prop_Jump_High", color='lightblue', ax=ax)

    # Scaffolding
    ax.set_title(f"Season {season}", weight='bold')
    ax.set_xlabel("Date", weight="bold")
    ax.set_ylabel('')
    ax.legend_.remove()
    sns.despine(ax=ax, top=True, right=True, left=False, bottom=False)
    ax.tick_params(axis='both', which='both', length=0, width=0)

    # Set x-axis ticks to selected dates
    ax.set_xticks(date_ticks)
    ax.set_xticklabels([dt.strftime('%b %Y') for dt in date_ticks])

# Adding title, subtitle, axis label, etc
fig.text(0.125, 1, 'High-Band Jump Trends by Player: Comparing', fontsize=20, rotation='horizontal', weight='bold')
fig.text(0.365, 1, 'One-Season', fontsize=20, rotation='horizontal', weight='bold', color='orange')
fig.text(0.429, 1, 'vs', fontsize=20, rotation='horizontal', weight='bold')
fig.text(0.443, 1, 'Two-Season', fontsize=20, rotation='horizontal', weight='bold', color='navy')
fig.text(0.506, 1, 'Tenure', fontsize=20, rotation='horizontal', weight='bold')
fig.text(0.125, 0.965, "No clear trend appears amongst one-season players, whereas two-season players show growth in their second season, as highlighted by", fontsize=12, ha="left")
fig.text(0.51, 0.965, "Athlete C", color = "lightblue", fontsize = 12)
fig.text(0.125, 0.9, 'High-Band Proportion', fontsize=10, rotation='horizontal', weight='bold')
plt.subplots_adjust(wspace=0.05)

# Saving and Showing Plot
plt.savefig("./images/viz_1.png", bbox_inches='tight')
plt.show()
```

## Making Visualization 2

### Creating "Adjusted_Day" Variable

```{python}
# Making a new variable which tells whether or not the player is in both seasons or just one
rolling_df['Player_Group'] = rolling_df['About'].apply(lambda x: 'both_seasons' if x in players_in_both else 'one_season')

# Getting the start date for each season
season_starts = rolling_df.groupby('Season')['Date'].min().to_dict()

# Getting the number of days into the season for each season
rolling_df['Day_of_Season'] = rolling_df.apply(lambda row: (row['Date'] - season_starts[row['Season']]).days + 1, axis=1)

# Getting last day of season 1
season1_length = (rolling_df[rolling_df['Season'] == 1]['Date'].max() - season_starts[1]).days + 1

# Function adds the number of days in the first season to the day in the second season
    # This prevents for example the 5th day in the second season and the 5th day in the first season being seen as the same day
def adjust_day(row):
    if row['Player_Group'] == 'both_seasons' and row['Season'] == 2:
        return row['Day_of_Season'] + season1_length
    else:
        return row['Day_of_Season']

# Adding the first season length to the second seasons
rolling_df['Adjusted_Day'] = rolling_df.apply(adjust_day, axis=1)
```

### Generating and Saving Visualization 2
```{python}
# Filter for Athlete C
Athlete_C = rolling_df[rolling_df["About"] == "Athlete C"]

# Plot both seasons on a single real-time timeline using Adjusted_Day
plt.figure(figsize=(14, 6))

sns.lineplot(data=Athlete_C, x='Adjusted_Day', y='Prop_Jump_High', marker='o', color='navy', label='High')
sns.lineplot(data=Athlete_C, x='Adjusted_Day', y='Prop_Jump_Low', marker='o', color='orange', label='Low')

# Adding line between Season 1 and Season 2
season1_len = (rolling_df[rolling_df['Season'] == 1]['Date'].max() - rolling_df[rolling_df['Season'] == 1]['Date'].min()).days + 1
plt.axvline(season1_len, linestyle='--', color='lightgray', label='Season Break')

# Adding subtitle
plt.suptitle('Athlete C Sees Significant Change in Jump-Band Proportions Over the Course of 2 Seasons', x = 0.33, y = 0.965, weight = "bold")

# Adding title and subtitle
plt.title("The two seasons Athlete C spends in the program results in notable growth in high-band jump proportion and decrease in the low-band jump proportion", ha = "left", fontsize = 10, x = 0, y = 1.075)
plt.text(276, 0.33, "High Band", color = "navy", weight = "bold")
plt.text(276, 0.3875, "Low Band", color = "orange", weight = "bold")
plt.text(60, 0.55, "Season 1", color = "gray", weight = "bold")
plt.text(180, 0.55, "Season 2", color = "gray", weight = "bold")
plt.xlabel("Adjusted Day", weight = "bold")
plt.text(-15, 0.625, "Proportion of Jumps", weight = "bold")
plt.ylabel("")

# Formatting
plt.tight_layout()
plt.gca().tick_params(axis = 'both', length = 0)
sns.despine(top = True, right = True)
plt.gca().get_legend().remove()

# Saving and Showing Plot
plt.savefig("./images/viz_2.png", bbox_inches='tight')
plt.show()
```

## Making Visualization 3

### Getting Daily Averages
```{python}
# Grouping by day of the season and whether or not the player was in both seasons, then taking the mean of the prop_jump_high
daily_means = rolling_df.groupby(['Adjusted_Day', 'Player_Group'])['Prop_Jump_High'].mean().reset_index()
```

### Getting Regression Lines for 1 and 2 Season Players
```{python}
# Making a dataframe to hold the required linear regression values
linreg_df = pd.DataFrame(columns=["Player_Group", "Slope", "Intercept", "R-value"])

# Looping through each unique player group
for group in daily_means["Player_Group"].unique():
    # Selecting the relevant data for the current player group
    group_data = daily_means[daily_means["Player_Group"] == group]
    
    # Running linear regression on Adjusted_Day vs Prop_Jump_High
    slope, intercept, r_value, p_value, std_err = stats.linregress(group_data['Adjusted_Day'], group_data['Prop_Jump_High'])
    
    # Adding the linear regression results to the DataFrame
    linreg_df.loc[len(linreg_df)] = [group, slope, intercept, r_value]

# Showing the table with linear regression results
linreg_df.head()
```

### Generating and Saving Visualization 3
```{python}
# Making palette for below plot
palette = {
    "both_seasons" : "navy",
    "one_season" : "orange"
}

# Making linear regression plot of prop_jump_high by # seasons played
plot = sns.lmplot(
    data = daily_means,
    x = 'Adjusted_Day',
    y = 'Prop_Jump_High',
    hue = 'Player_Group',
    palette = palette,
    markers = ["o", "s"],
    scatter_kws = {"s": 50, "alpha": 0.5},
    height = 6,
    aspect = 2
)

# Adding subtitle and title
plt.suptitle('Two-Season Players See Positive Growth in High-Band Jump Proportions'
             , x = 0.305, y = 1.125, weight = "bold")
plt.title("On average, players with 2 seasons of data increased their proportion of jumps in the high band, while 1-season players\nsaw an average decrease (likely due to fatigue, not regression).", x = 0, y = 1.055, ha = "left", fontsize = 10)

# Scaffolding/Formatting
plt.xlabel('Adjusted Day', weight = 'bold')
plt.text(-15, 0.32, "Mean Proportion of Jumps in High Band", weight = "bold")
plt.ylabel('')
plt.text(275, 0.259, "Two Seasons", color = "navy", weight = 'bold')
plt.text(150, 0.195, "One Season", color = "orange", weight = 'bold')
plt.gca().tick_params(axis = 'both', length = 0)
plot._legend.remove()

# Saving and Showing Plot
plt.savefig("./images/viz_3.png", bbox_inches='tight')
plt.show()
```