Chapter 14: Key Takeaways - Player and Team Comparison Charts

Quick Reference Card

Comparison Principles

Chart Type Selection Guide

When to Use Each Chart

Code Patterns

Radar Chart Template

def create_radar(values, metrics, title):
    angles = np.linspace(0, 2*np.pi, len(metrics), endpoint=False).tolist()
    angles += angles[:1]
    values += values[:1]

    fig, ax = plt.subplots(subplot_kw=dict(polar=True))
    ax.plot(angles, values, 'o-', linewidth=2)
    ax.fill(angles, values, alpha=0.25)
    ax.set_xticks(angles[:-1])
    ax.set_xticklabels(metrics)
    ax.set_ylim(0, 1)
    return fig

Dumbbell Chart Template

def create_dumbbell(entities, val1, val2, labels):
    fig, ax = plt.subplots()
    for y, (v1, v2) in enumerate(zip(val1, val2)):
        color = '#2a9d8f' if v2 > v1 else '#e76f51'
        ax.plot([v1, v2], [y, y], color=color, linewidth=2)
        ax.scatter([v1], [y], s=80, color='#264653')
        ax.scatter([v2], [y], s=80, color='#f4a261')
    ax.set_yticks(range(len(entities)))
    ax.set_yticklabels(entities)
    ax.invert_yaxis()
    return fig

Bump Chart Template

def create_bump(rankings, periods, highlight=None):
    fig, ax = plt.subplots()
    for entity, ranks in rankings.items():
        alpha = 1.0 if highlight and entity in highlight else 0.3
        ax.plot(range(len(periods)), ranks, 'o-', alpha=alpha)
    ax.set_xticks(range(len(periods)))
    ax.set_xticklabels(periods)
    ax.invert_yaxis()  # Rank 1 at top
    return fig

Percentile Color Scale

def get_percentile_color(pct):
    if pct >= 75: return '#1a9641'    # Elite (green)
    elif pct >= 50: return '#a6d96a'  # Above avg (light green)
    elif pct >= 25: return '#ffffbf'  # Average (yellow)
    elif pct >= 10: return '#fdae61'  # Below avg (orange)
    else: return '#d7191c'            # Poor (red)

Fairness Checklist

Before publishing any comparison:

• [ ] Position equivalence: Are compared players at same position/role?
• [ ] Opportunity adjustment: Normalized for games played, attempts, snaps?
• [ ] Era adjustment: Accounted for rule/strategy changes if comparing across years?
• [ ] Metric balance: Selected metrics fairly represent all compared entities?
• [ ] Context provided: Includes percentiles, averages, or benchmarks?
• [ ] Scale appropriate: Axes start at sensible values, not truncated to exaggerate?
• [ ] Methodology disclosed: How were similarities/rankings calculated?

Normalization Formulas

Min-Max Normalization (0-1 Scale)

normalized = (value - min) / (max - min)

Z-Score Standardization

z = (value - mean) / std_dev

Percentile Rank

from scipy.stats import percentileofscore
pct = percentileofscore(population, value)

Inverted Metrics (lower is better)

inverted_normalized = 1 - (value - min) / (max - min)

Color Conventions

Performance Encoding

Percentile Zones

Common Mistakes to Avoid

1. Cherry-Picking Metrics

Wrong: Include only passing metrics when comparing a better passer to a better runner Right: Include balanced set of relevant position metrics

2. Inconsistent Baselines

Wrong: Start one bar at 0, another at 50 Right: All bars start at same baseline

3. Missing Context

Wrong: "Player has 68% completion rate" Right: "Player has 68% completion rate (42nd percentile)"

4. Radar Chart Overload

Wrong: 15 metrics crammed into radar chart Right: 5-8 carefully selected metrics

5. Ignoring Opportunity

Wrong: Compare backup's per-play stats to starter's Right: Note sample size and game situations

Similarity Calculation

Euclidean Distance

from scipy.spatial.distance import euclidean
dist = euclidean(player_a_features, player_b_features)
similarity = 1 / (1 + dist)

Cosine Similarity

from numpy import dot
from numpy.linalg import norm
similarity = dot(a, b) / (norm(a) * norm(b))

Hierarchical Clustering

from scipy.cluster.hierarchy import linkage, dendrogram
from sklearn.preprocessing import StandardScaler

scaled = StandardScaler().fit_transform(features)
linkage_matrix = linkage(scaled, method='ward')

Quick Decision Tree

What type of comparison?
│
├── Single metric, many entities → Horizontal Bar Chart
│
├── Multiple metrics, few entities (2-4) → Grouped Bar or Radar
│
├── Two time points → Dumbbell or Slope Chart
│
├── Rankings over time → Bump Chart
│
├── Same analysis, many groups → Small Multiples
│
├── Finding similar entities → Heatmap + Dendrogram
│
└── Value in context → Percentile Chart

Key Terminology