WHOOP Data Analysis – Luke Muoio & Will Kang

Loading data...

\n\n

About This Project

This dashboard analyzes real WHOOP fitness tracker data collected from Luke Muoio, his friends, and family members, making it a multi-person dataset rather than a single individual's log.
The WHOOP is a wearable that continuously measures strain, recovery, sleep quality, and heart rate variability (HRV) to give athletes insight into how their body is responding to daily life.
The goal of this project was to move beyond the WHOOP app's built-in summaries and ask deeper questions: What actually drives recovery? How does sleep composition change with total sleep time? Do lifestyle patterns like weekend behavior show up in the data?
This analysis was built for IST 462 (Scripting for Data Analysis) at Syracuse University, with all charts reproduced from the original Python/Colab notebook.

\n\n

Statistical Distributions of All WHOOP Tracking Categories

These 20 histograms give a baseline picture of each tracked metric across the full dataset. The KDE curve (red line) smooths out the shape so you can see whether values cluster tightly or spread widely.
Metrics like blood oxygen % and sleep efficiency cluster near their upper limits, while recovery score and HRV show much wider and flatter distributions, which shows just how much these numbers move around from day to day.

\n\n

View Python code

# Columns to analyze\ncolumns_to_analyze = ['Recovery score %', 'Resting heart rate (bpm)', 'Skin temp (celsius)', 'Blood oxygen %', 'Day Strain', 'Energy burned (cal)',\n                       'Max HR (bpm)', 'Average HR (bpm)', 'Sleep performance %', 'Respiratory rate (rpm)', 'Asleep duration (min)',\n                       'In bed duration (min)', 'Light sleep duration (min)', 'Deep (SWS) duration (min)', 'REM duration (min)',\n                       'Awake duration (min)', 'Sleep need (min)', 'Sleep debt (min)', 'Sleep efficiency %', 'Sleep consistency %']\n\n# Create subplots\nfig, axes = plt.subplots(nrows=7, ncols=3, figsize=(15, 20))\nfig.subplots_adjust(hspace=0.5)\n\n# Plot histograms for each column\nfor i, column in enumerate(columns_to_analyze): # enumerate boss so boss luke look at this its so boss\n    if column in whoop_df.columns:\n        sns.histplot(whoop_df[column].dropna(), kde=True, ax=axes[i // 3, i % 3])\n        axes[i // 3, i % 3].set_title(f'Distribution of {column}')\n    else:\n        print(f"Column '{column}' not found in the dataframe.")\n\n# Show the plot\nplt.show()

\n\n

Light Sleep vs. Total Sleep (2022) – Bubble Size: Recovery Score

With more total sleep, light sleep consistently grows the most relative to deep or REM sleep, suggesting the body fills additional time in bed with lighter rest rather than proportionally more restorative stages.
Bubble size encodes recovery score. Notice that higher-recovery nights don't always correspond to the longest sleep, pointing to quality over quantity.

\n\n

View Python code

whoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\nwhoop_2023 = whoop_df[(whoop_df['Cycle start time'].dt.year == 2022)]\n\nplt.scatter(whoop_2023['Asleep duration (min)'],whoop_2023['Light sleep duration (min)'], s=whoop_2023['Recovery score %'], alpha=0.3, c='blue', label='Bubble Chart')\n\n# Add labels and title\nplt.xlabel('Time Spent Asleep (min)')\nplt.ylabel('Time Spent in Light sleep (min)')\nplt.title('Relationship between Time Spent in Light Sleep vs. Total Sleep (2022)')\n\n\n# Show the plot\nplt.show()

\n\n

Heart Rate Variability's Effect on Recovery Score

HRV is highly individual: a score of 70ms might indicate excellent recovery for one person and poor recovery for another, which is why the data appears as two distinct bands rather than a single clean trend.
WHOOP's recovery algorithm is calibrated to each user's personal HRV baseline, so this chart is best read as \"relative to yourself\" rather than across people.

\n\n

View Python code

whoop_df.plot(x='Heart Rate variability (ms)', y='Recovery score %', kind='scatter')\nplt.title('Heart rate variabillity\\\\'s effect on recovery score')\nplt.xlabel('Heart rate variabillity (ms)')\nplt.ylabel('Recovery Score (%)')\nplt.show()

\n\n

Energy Burned vs. Mean Average Heart Rate (with Std Dev)

As energy expenditure climbs, average heart rate rises in step, but the error bars widen considerably at higher output, reflecting how differently people's hearts respond to intense effort.
The categories cut off at 3,000 cal because most days in the dataset fell below that threshold.

\n\n

View Python code

# Binning 'Energy burned' with extended categories\nbins = np.arange(1000, 3200, 200)\nlabels = [f'{i}-{i+200}' for i in range(1000, 3000, 200)]\n\nwhoop_df['Energy burned category'] = pd.cut(whoop_df['Energy burned (cal)'], bins=bins, labels=labels)\n\n# Create a bar graph with error bars\nplt.figure(figsize=(10, 6))\nbar_data = whoop_df.groupby('Energy burned category')['Average HR (bpm)'].agg(['mean', 'std'])\nbar_data['mean'].plot(kind='bar', yerr=bar_data['std'], capsize=5, color='skyblue')\n\nplt.title('Bar Graph with Error Bars: Energy Burned vs. Mean Average Heart Rate')\nplt.xlabel('Energy Burned Category')\nplt.ylabel('Mean Average Heart Rate')\nplt.show()

\n\n

Daily Strain Distribution by Day of the Week

Strain is distributed fairly evenly Monday through Friday, with a slight uptick on Thursdays and Fridays, consistent with people pushing harder at the end of the workweek.
Weekends show a noticeable dip in high-strain days, which aligns with typical rest and recovery behavior.

\n\n

View Python code

# Assuming your DataFrame is named 'whoop_df'\n# Assuming 'Cycle start time' is in datetime format\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\n\n# Extract day of the week from 'Cycle start time'\nwhoop_df['Day of Week'] = whoop_df['Cycle start time'].dt.day_name()\n\n# Define the order of days\ndays_order = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']\n\n# Create a swarm plot with custom order\nplt.figure(figsize=(20, 8))\nsns.swarmplot(x='Day of Week', y='Day Strain', data=whoop_df, palette='viridis', order=days_order)\nplt.title('Daily Strain Distribution by Day of the Week')\nplt.xlabel('Day of the Week')\nplt.ylabel('Daily Strain')\nplt.show()

\n\n

Top 10 Highest Strain Days

All 10 of the highest-strain days exceeded a score of 18 out of 21. WHOOP considers anything above 18 to be \"all out\" exertion.
The y-axis starts at 18 rather than zero to highlight the narrow range these peak days occupy. Every day on this chart is a genuine outlier in the dataset.

\n\n

View Python code

# Identify the top 10 worst strain scores\ntop_10_worst_strain = whoop_df.nlargest(10, 'Day Strain')\n\n# Plot the top 10 worst strain scores\nplt.figure(figsize=(12, 8))\nplt.bar(top_10_worst_strain['Cycle start time'], top_10_worst_strain['Day Strain'], color='red', alpha=0.7)\n\n# Set the y-axis limits\nplt.ylim(18, 21)\n\nplt.title('Top 10 Worst Strain Scores')\nplt.xlabel('Cycle start time')\nplt.ylabel('Day Strain')\nplt.xticks(rotation=45, ha='right')  # Rotate x-axis labels for better readability\nplt.show()

\n\n

Skin Temp & Blood Oxygen vs. REM Duration (Regression)

Higher skin temperature during sleep is associated with shorter REM duration. One practical takeaway is that sleeping in a cooler room may help extend the most cognitively restorative sleep stage.
Blood oxygen showed the opposite pattern: higher SpO2 correlated with more REM sleep, which makes physiological sense as the brain is highly active during REM and depends on good oxygenation.

\n\n

View Python code

# Assuming 'Skin temp (celsius)', 'Blood oxygen %', and 'REM duration (min)' columns exist in whoop_df\nwhoop_df['Skin temp (celsius)'] = pd.to_numeric(whoop_df['Skin temp (celsius)'], errors='coerce')\nwhoop_df['Blood oxygen %'] = pd.to_numeric(whoop_df['Blood oxygen %'], errors='coerce')\nwhoop_df['REM duration (min)'] = pd.to_numeric(whoop_df['REM duration (min)'], errors='coerce')\n\n# Drop rows with NaN values in the specified columns\nwhoop_df = whoop_df.dropna(subset=['Skin temp (celsius)', 'Blood oxygen %', 'REM duration (min)'])\n\n# Create scatter plots with red regression lines\nplt.figure(figsize=(15, 8))\n\n# Scatter plot for 'Skin temp (celsius)' vs. 'REM duration (min)' with red regression line\nplt.subplot(2, 1, 1)\nsns.regplot(x='Skin temp (celsius)', y='REM duration (min)', data=whoop_df, scatter_kws={'alpha':0.5}, line_kws={'color': 'red'})\nplt.title('Skin Temp vs. REM Duration')\nplt.xlabel('Skin Temp (Celsius)')\nplt.ylabel('REM Duration (min)')\nplt.grid(True)\n\n# Scatter plot for 'Blood oxygen %' vs. 'REM duration (min)' with red regression line\nplt.subplot(2, 1, 2)\nsns.regplot(x='Blood oxygen %', y='REM duration (min)', data=whoop_df, scatter_kws={'alpha':0.5}, line_kws={'color': 'red'})\nplt.title('Blood Oxygen % vs. REM Duration')\nplt.xlabel('Blood Oxygen %')\nplt.ylabel('REM Duration (min)')\nplt.grid(True)\n\nplt.tight_layout()\nplt.show()

\n\n

Sleep Metrics Over Time – 2022 (Stack Plot)

Across 2022, awake time stays relatively stable while REM and deep sleep fluctuate more, suggesting the body prioritizes staying asleep even when sleep quality varies.

\n\n

View Python code

# Assuming 'Cycle start time' is in datetime format\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\n\n# Filter DataFrame to include only data from the year 2022\nwhoop_df_2022 = whoop_df[(whoop_df['Cycle start time'] >= '2022-01-01') & (whoop_df['Cycle start time'] < '2023-01-01')]\n\n# Sort DataFrame by 'Cycle start time' for a chronological line plot\nwhoop_df_2022 = whoop_df_2022.sort_values(by='Cycle start time')\n\n# Create a stack plot for different sleep metrics over time in 2022\nplt.figure(figsize=(15, 8))\n\nplt.stackplot(whoop_df_2022['Cycle start time'],\n              whoop_df_2022['REM duration (min)'],\n              whoop_df_2022['Deep (SWS) duration (min)'],\n              whoop_df_2022['Awake duration (min)'],\n              labels=['REM Duration (min)', 'Deep Sleep Duration (min)', 'Awake Duration (min)'])\n\nplt.title('Sleep Metrics Over Time in 2022 (Stack Plot)')\nplt.xlabel('Cycle start time')\nplt.ylabel('Duration (min)')\nplt.legend()\nplt.grid(True)\nplt.show()

\n\n

Sleep Metrics – September 2022 (Stack Plot)

The September view was specifically chosen because it captures the start of the academic year. Syllabus week and early-semester patterns are visible in the data as sleep metrics shift.

\n\n

View Python code

# Assuming 'Cycle start time' is in datetime format\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\n\n# Filter DataFrame to include only data from the year 2022\nwhoop_df_2022 = whoop_df[(whoop_df['Cycle start time'] >= '2022-09-01') & (whoop_df['Cycle start time'] < '2022-10-01')]\n\n# Sort DataFrame by 'Cycle start time' for a chronological line plot\nwhoop_df_2022 = whoop_df_2022.sort_values(by='Cycle start time')\n\n# Create a stack plot for different sleep metrics over time in 2022\nplt.figure(figsize=(12, 8))\n\nplt.stackplot(whoop_df_2022['Cycle start time'],\n              whoop_df_2022['REM duration (min)'],\n              whoop_df_2022['Deep (SWS) duration (min)'],\n              whoop_df_2022['Awake duration (min)'],\n              labels=['REM Duration (min)', 'Deep Sleep Duration (min)', 'Awake Duration (min)'])\n\nplt.title('Sleep Metrics Over Time in the month of September')\nplt.xlabel('Cycle start time')\nplt.ylabel('Duration (min)')\nplt.legend()\nplt.grid(True)\nplt.show()

\n\n

Recovery Score – September 2023 (Weekends Marked in Red)

The red dots (Fridays and Saturdays) cluster consistently at lower recovery values throughout the month. The original notebook notes this is most likely explained by elevated alcohol consumption on weekends.
This was one of the most statistically clear findings in the dataset: weekend behavior leaves a measurable physiological signature the following morning.

\n\n

View Python code

# Assuming 'Cycle start time' is in datetime format\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\n\n# Filter DataFrame to include only data from September 2023\nseptember_2023_data = whoop_df[(whoop_df['Cycle start time'].dt.year == 2023) & (whoop_df['Cycle start time'].dt.month == 9)]\n\n# Sort DataFrame by 'Cycle start time' for a chronological line plot\nseptember_2023_data = september_2023_data.sort_values(by='Cycle start time')\n\n# Create a line plot for recovery over the month of September 2023\nplt.figure(figsize=(12, 8))\n\n# Plot the recovery scores\nplt.plot(september_2023_data['Cycle start time'], september_2023_data['Recovery score %'], linestyle='-', color='blue', label='Recovery Score %')\n\n# Identify and mark weekends with red dots\nweekends = september_2023_data[september_2023_data['Cycle start time'].dt.day_name().isin(['Friday', 'Saturday'])]\nplt.scatter(weekends['Cycle start time'], weekends['Recovery score %'], color='red', label='Weekends')\n\n# Set labels and title\nplt.title('Recovery Over the Month of September 2023 with Weekends Marked')\nplt.xlabel('Date')\nplt.ylabel('Recovery Score %')\n\n# Add legend\nplt.legend()\n\n# Add grid\nplt.grid(True)\n\n# Show the plot\nplt.show()

\n\n

Average Recovery Score by Day of the Week

Monday and Tuesday show some of the highest average recovery scores, likely because the body has had the weekend to rest from physical strain, even when sleep quality takes a hit on those same nights.
Friday and Saturday sit at the bottom, connecting directly to the September recovery chart above.

\n\n

View Python code

# Assuming 'Cycle start time' and 'Recovery score %' columns exist in whoop_df\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\nwhoop_df['Day of Week'] = whoop_df['Cycle start time'].dt.day_name()\n\n# Group by day of the week and calculate the mean recovery score\naverage_recovery_by_day = whoop_df.groupby('Day of Week')['Recovery score %'].mean()\n\n# Order the days of the week\ndays_order = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']\n\n# Create a bar plot\nplt.figure(figsize=(10, 6))\nax = average_recovery_by_day.loc[days_order].plot(kind='bar', color='green')\nplt.title('Average Recovery Score by Day of the Week')\nplt.xlabel('Day of the Week')\nplt.ylabel('Average Recovery Score %')\n\n# Add a line across the top of each bar\nfor p in ax.patches:\n    ax.annotate(str(round(p.get_height(), 2)), (p.get_x() + p.get_width() / 2., p.get_height()),\n                ha='center', va='baseline', xytext=(0, 5), textcoords='offset points')\n\nplt.show()

\n\n

Distribution of Recovery Scores (All vs. Thu–Sat)

The orange overlay (Thursday through Saturday) shifts the distribution noticeably leftward, confirming that late-week and weekend nights consistently produce lower recovery scores. This holds across the full dataset, not only during September 2023.
The distribution for all scores is roughly bell-shaped, meaning extreme highs and lows are equally rare.

\n\n

View Python code

# Assuming 'Cycle start time' and 'Recovery score %' columns exist in whoop_df\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\nwhoop_df['Day of Week'] = whoop_df['Cycle start time'].dt.day_name()\n\n# Filter recovery scores on weekends\nweekend_scores = whoop_df.loc[whoop_df['Day of Week'].isin(['Thursday', 'Friday', 'Saturday']), 'Recovery score %'].dropna()\n\n# Create a histogram with kernel density estimate\nplt.figure(figsize=(10, 6))\nax = sns.histplot(whoop_df['Recovery score %'].dropna(), bins=20, kde=True, color='skyblue', edgecolor='black', label='All Scores')\nsns.histplot(weekend_scores, bins=20, kde=True, color='orange', alpha=0.3, label='Thursday-Saturday')\n\nplt.title('Distribution of Recovery Scores')\nplt.xlabel('Recovery Score %')\nplt.ylabel('Frequency')\nplt.grid(True)\nplt.legend()\n\nplt.show()\n\n# Drop rows with missing values in the relevant columns\nsleep_recovery_data = whoop_df[['Sleep consistency %', 'Recovery score %']].dropna()\n\n# Scatter plot with green color\nplt.figure(figsize=(10, 6))\nsns.regplot(x='Sleep consistency %', y='Recovery score %', data=sleep_recovery_data, scatter_kws={'alpha': 0.5, 'color': 'green'}, line_kws={'color': 'purple'})\nplt.title('Scatter Plot: Sleep Consistency vs. Recovery Score')\nplt.xlabel('Sleep Consistency %')\nplt.ylabel('Recovery Score %')\nplt.grid(True)\nplt.show()\n\n# Calculate correlation coefficient\ncorrelation_coefficient = sleep_recovery_data['Sleep consistency %'].corr(sleep_recovery_data['Recovery score %'])\nprint(f'Correlation Coefficient: {correlation_coefficient}')

\n\n

Sleep Consistency vs. Recovery Score

Going to sleep and waking up at consistent times has a modest positive correlation with recovery. WHOOP actively rewards sleep consistency in its algorithm.
The scatter is wide, which reflects that consistency is just one of many factors. Someone who sleeps inconsistently but deeply may still recover well.

\n\n

View Python code

# Assuming 'Cycle start time' and 'Recovery score %' columns exist in whoop_df\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\nwhoop_df['Day of Week'] = whoop_df['Cycle start time'].dt.day_name()\n\n# Filter recovery scores on weekends\nweekend_scores = whoop_df.loc[whoop_df['Day of Week'].isin(['Thursday', 'Friday', 'Saturday']), 'Recovery score %'].dropna()\n\n# Create a histogram with kernel density estimate\nplt.figure(figsize=(10, 6))\nax = sns.histplot(whoop_df['Recovery score %'].dropna(), bins=20, kde=True, color='skyblue', edgecolor='black', label='All Scores')\nsns.histplot(weekend_scores, bins=20, kde=True, color='orange', alpha=0.3, label='Thursday-Saturday')\n\nplt.title('Distribution of Recovery Scores')\nplt.xlabel('Recovery Score %')\nplt.ylabel('Frequency')\nplt.grid(True)\nplt.legend()\n\nplt.show()\n\n# Drop rows with missing values in the relevant columns\nsleep_recovery_data = whoop_df[['Sleep consistency %', 'Recovery score %']].dropna()\n\n# Scatter plot with green color\nplt.figure(figsize=(10, 6))\nsns.regplot(x='Sleep consistency %', y='Recovery score %', data=sleep_recovery_data, scatter_kws={'alpha': 0.5, 'color': 'green'}, line_kws={'color': 'purple'})\nplt.title('Scatter Plot: Sleep Consistency vs. Recovery Score')\nplt.xlabel('Sleep Consistency %')\nplt.ylabel('Recovery Score %')\nplt.grid(True)\nplt.show()\n\n# Calculate correlation coefficient\ncorrelation_coefficient = sleep_recovery_data['Sleep consistency %'].corr(sleep_recovery_data['Recovery score %'])\nprint(f'Correlation Coefficient: {correlation_coefficient}')

\n\n

Distribution of Heart Rate Variability

HRV distribution is right-skewed, with most readings clustering at lower values and a long tail of high-HRV days that represent peak recovery.
Because the dataset pools multiple people, this distribution reflects both individual baselines and day-to-day fluctuation within each person.

\n\n

View Python code

# Drop rows with missing values in the relevant column\ndata = whoop_df.dropna(subset=['Heart rate variability (ms)'])\n\n# Create a histogram and kernel density plot\nplt.figure(figsize=(10, 6))\nsns.histplot(data['Heart rate variability (ms)'], kde=True, color='skyblue')\nplt.title('Distribution of Heart Rate Variability')\nplt.xlabel('Heart Rate Variability (ms)')\nplt.ylabel('Frequency in Dataframe')\nplt.show()\n\n# Graph is a positive skew

\n\n

Skin Temp vs. Heart Rate Variability (with Trend Line & Anomalies)

Higher skin temperature correlates with lower HRV, which in turn drives lower recovery scores. A warmer body at night is a signal that the body is still in a stressed or inflammatory state.
The highlighted anomalies (triangle markers) are days where HRV deviated more than 2 standard deviations from what the skin temp trend would predict. These could represent illness, unusually hard training, or other acute stressors.

\n\n

View Python code

# Drop rows with missing values in relevant columns\ndata = whoop_df.dropna(subset=['Skin temp (celsius)', 'Heart rate variability (ms)'])\n\n# Model the relationship using linear regression\nX = sm.add_constant(data['Skin temp (celsius)'])\ny = data['Heart rate variability (ms)']\n\nmodel = sm.OLS(y, X).fit()\n\n# Calculate residuals\nresiduals = y - model.predict(X)\n\n# Set the threshold for anomalies in residuals (e.g., 2 standard deviations)\nthreshold_std_residuals = 2\n\n# Identify anomalies in residuals\nanomalies_residuals = data[abs(residuals) > threshold_std_residuals * residuals.std()]\n\n# Create the line graph with regression line\nplt.figure(figsize=(10, 8))\nsns.lineplot(data=data, x='Skin temp (celsius)', y='Heart rate variability (ms)', label='Heart Rate Variability', color='green')\nplt.plot(data['Skin temp (celsius)'], model.predict(X), color='lightgreen', label='Trend Line')\n\n# Highlight anomalies in residuals\nif not anomalies_residuals.empty:\n    plt.scatter(anomalies_residuals['Skin temp (celsius)'], anomalies_residuals['Heart rate variability (ms)'], color='lightgreen', label='Anomalies')\n\nplt.title('Line Graph: Skin Temp vs. Heart Rate Variability with Anomalies')\nplt.xlabel('Skin Temp (Celsius)')\nplt.ylabel('Heart Rate Variability (ms)')\nplt.legend()\nplt.show()

\n\n

Heart Rate Metrics Over Time (Bubble: size & color = HRV)

Each bubble plots resting heart rate over time, with bubble size and color both encoding HRV. Larger, brighter bubbles represent days with high HRV and low resting HR, the combination WHOOP considers optimal recovery.
Reading left to right across time, you can see periods where both metrics improve together (bigger, brighter bubbles) and periods of stress where small, dark bubbles cluster.

\n\n

View Python code

# Assuming 'Cycle start time' is in datetime format\nwhoop_df['Cycle start time'] = pd.to_datetime(whoop_df['Cycle start time'])\n\n# Filter data for the year 2022 and onwards\nwhoop_df = whoop_df[whoop_df['Cycle start time'].dt.year >= 2022]\n\n# Sort DataFrame by 'Cycle start time' for a chronological bubble plot\nwhoop_df = whoop_df.sort_values(by='Cycle start time')\n\n# Create a bubble chart for heart rate metrics over time\nplt.figure(figsize=(12, 8))\n\nplt.scatter(whoop_df['Cycle start time'],\n            whoop_df['Resting heart rate (bpm)'],\n            s=whoop_df['Heart rate variability (ms)'],  # Map size to frequency of heart rate variability\n            c=whoop_df['Heart rate variability (ms)'],  # Map color to heart rate variability\n            alpha=0.7,\n            cmap='viridis',\n            edgecolors='w',\n            linewidth=0.5)\n\nplt.title('Heart Rate Metrics Over Time (Bubble Chart)')\nplt.xlabel('Cycle start time')\nplt.ylabel('Resting Heart Rate (bpm)')\nplt.colorbar(label='Heart Rate Variability (ms)')  # Add colorbar to indicate bubble sizes\nplt.grid(True)\nplt.show()

\n\n

Loading data...

\n\n

About This Project

This dashboard analyzes real WHOOP fitness tracker data collected from Luke Muoio, his friends, and family members, making it a multi-person dataset rather than a single individual's log.
The WHOOP is a wearable that continuously measures strain, recovery, sleep quality, and heart rate variability (HRV) to give athletes insight into how their body is responding to daily life.
The goal of this project was to move beyond the WHOOP app's built-in summaries and ask deeper questions: What actually drives recovery? How does sleep composition change with total sleep time? Do lifestyle patterns like weekend behavior show up in the data?
This analysis was built for IST 462 (Scripting for Data Analysis) at Syracuse University, with all charts reproduced from the original Python/Colab notebook.

\n\n

Statistical Distributions of All WHOOP Tracking Categories

These 20 histograms give a baseline picture of each tracked metric across the full dataset. The KDE curve (red line) smooths out the shape so you can see whether values cluster tightly or spread widely.
Metrics like blood oxygen % and sleep efficiency cluster near their upper limits, while recovery score and HRV show much wider and flatter distributions, which shows just how much these numbers move around from day to day.

\n\n

Light Sleep vs. Total Sleep (2022) – Bubble Size: Recovery Score

With more total sleep, light sleep consistently grows the most relative to deep or REM sleep, suggesting the body fills additional time in bed with lighter rest rather than proportionally more restorative stages.
Bubble size encodes recovery score. Notice that higher-recovery nights don't always correspond to the longest sleep, pointing to quality over quantity.

\n\n

Heart Rate Variability's Effect on Recovery Score

HRV is highly individual: a score of 70ms might indicate excellent recovery for one person and poor recovery for another, which is why the data appears as two distinct bands rather than a single clean trend.
WHOOP's recovery algorithm is calibrated to each user's personal HRV baseline, so this chart is best read as \"relative to yourself\" rather than across people.

\n\n

Energy Burned vs. Mean Average Heart Rate (with Std Dev)

As energy expenditure climbs, average heart rate rises in step, but the error bars widen considerably at higher output, reflecting how differently people's hearts respond to intense effort.
The categories cut off at 3,000 cal because most days in the dataset fell below that threshold.

\n\n

Daily Strain Distribution by Day of the Week

Strain is distributed fairly evenly Monday through Friday, with a slight uptick on Thursdays and Fridays, consistent with people pushing harder at the end of the workweek.
Weekends show a noticeable dip in high-strain days, which aligns with typical rest and recovery behavior.

\n\n

Top 10 Highest Strain Days

All 10 of the highest-strain days exceeded a score of 18 out of 21. WHOOP considers anything above 18 to be \"all out\" exertion.
The y-axis starts at 18 rather than zero to highlight the narrow range these peak days occupy. Every day on this chart is a genuine outlier in the dataset.

\n\n

Skin Temp & Blood Oxygen vs. REM Duration (Regression)

Higher skin temperature during sleep is associated with shorter REM duration. One practical takeaway is that sleeping in a cooler room may help extend the most cognitively restorative sleep stage.
Blood oxygen showed the opposite pattern: higher SpO2 correlated with more REM sleep, which makes physiological sense as the brain is highly active during REM and depends on good oxygenation.

\n\n

Sleep Metrics Over Time – 2022 (Stack Plot)

Across 2022, awake time stays relatively stable while REM and deep sleep fluctuate more, suggesting the body prioritizes staying asleep even when sleep quality varies.

\n\n

Sleep Metrics – September 2022 (Stack Plot)

The September view was specifically chosen because it captures the start of the academic year. Syllabus week and early-semester patterns are visible in the data as sleep metrics shift.

\n\n

Recovery Score – September 2023 (Weekends Marked in Red)

The red dots (Fridays and Saturdays) cluster consistently at lower recovery values throughout the month. The original notebook notes this is most likely explained by elevated alcohol consumption on weekends.
This was one of the most statistically clear findings in the dataset: weekend behavior leaves a measurable physiological signature the following morning.

\n\n

Average Recovery Score by Day of the Week

Monday and Tuesday show some of the highest average recovery scores, likely because the body has had the weekend to rest from physical strain, even when sleep quality takes a hit on those same nights.
Friday and Saturday sit at the bottom, connecting directly to the September recovery chart above.

\n\n

Distribution of Recovery Scores (All vs. Thu–Sat)

The orange overlay (Thursday through Saturday) shifts the distribution noticeably leftward, confirming that late-week and weekend nights consistently produce lower recovery scores. This holds across the full dataset, not only during September 2023.
The distribution for all scores is roughly bell-shaped, meaning extreme highs and lows are equally rare.

\n\n

Sleep Consistency vs. Recovery Score

Going to sleep and waking up at consistent times has a modest positive correlation with recovery. WHOOP actively rewards sleep consistency in its algorithm.
The scatter is wide, which reflects that consistency is just one of many factors. Someone who sleeps inconsistently but deeply may still recover well.

\n\n

Distribution of Heart Rate Variability

HRV distribution is right-skewed, with most readings clustering at lower values and a long tail of high-HRV days that represent peak recovery.
Because the dataset pools multiple people, this distribution reflects both individual baselines and day-to-day fluctuation within each person.

\n\n

Skin Temp vs. Heart Rate Variability (with Trend Line & Anomalies)

Higher skin temperature correlates with lower HRV, which in turn drives lower recovery scores. A warmer body at night is a signal that the body is still in a stressed or inflammatory state.
The highlighted anomalies (triangle markers) are days where HRV deviated more than 2 standard deviations from what the skin temp trend would predict. These could represent illness, unusually hard training, or other acute stressors.

\n\n

Heart Rate Metrics Over Time (Bubble: size & color = HRV)

Each bubble plots resting heart rate over time, with bubble size and color both encoding HRV. Larger, brighter bubbles represent days with high HRV and low resting HR, the combination WHOOP considers optimal recovery.
Reading left to right across time, you can see periods where both metrics improve together (bigger, brighter bubbles) and periods of stress where small, dark bubbles cluster.

\n\n

Diff · v24