Pie and Donut Charts: Showing Part-to-Whole (and When Not To)

I'll defend the pie chart, but only just. They earn their bad reputation most of the time — yet they do one job genuinely well: showing what share of a whole each part takes. Give a pie (or its hollow cousin, the donut) a few categories that sum to a meaningful total and it reads instantly. I'll build both for each division's share of league-wide runs, then show you the case where a plain bar is the honest choice instead.

This builds on Your First Visualization. The data is the bundled sample_standings.csv (real 2023 MLB standings), so it runs offline.

1. A real part-to-whole

   Pies only make sense when the slices genuinely sum to a whole. "Each division's share of all runs scored" qualifies: every run belongs to exactly one division, and the six shares add to 100%.

   import pandas as pd
   
   df = pd.read_csv("sample_standings.csv")
   runs_by_div = df.groupby("Division")["RS"].sum().sort_values(ascending=False)
   shares = (runs_by_div / runs_by_div.sum() * 100).round(1)
   print(shares.to_string())

   Each division's share of league-wide runs

   Share of league-wide runs scored, by division:
   Division
   AL East       17.2
   NL East       17.1
   NL West       16.9
   AL West       16.9
   NL Central    16.7
   AL Central    15.2

   Already a warning sign: the shares are all bunched between ~15% and ~17%. Hold that thought — it's exactly the pattern pies handle badly, and we'll come back to it.

2. Pie and donut, side by side

   ax.pie() takes the values and draws the slices; autopct prints each percentage. A donut is just a pie with a hole, set by wedgeprops=dict(width=...) — the hole is handy for a title or total in the center, and many find it easier to read.

   import matplotlib.pyplot as plt
   
   colors = plt.cm.Set2(range(len(runs_by_div)))
   fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(11, 5.5))
   
   ax1.pie(runs_by_div, labels=runs_by_div.index, colors=colors,
           autopct="%1.1f%%", startangle=90, counterclock=False)
   
   ax2.pie(runs_by_div, labels=runs_by_div.index, colors=colors,
           autopct="%1.1f%%", startangle=90, counterclock=False,
           wedgeprops=dict(width=0.42))          # the hole makes it a donut
   fig.savefig("pie_donut.png", dpi=144, bbox_inches="tight")

   

   startangle=90, counterclock=False starts the first slice at the top and goes clockwise — the order people instinctively read. Without it, pies start at 3 o'clock and run counterclockwise, which feels subtly wrong.

3. When a bar wins

   Look at the chart: six slices all near one-sixth. Can you tell, by eye, whether the AL East or NL East slice is bigger? Almost nobody can — the human eye judges lengths far better than angles or areas. That's the pie's fatal flaw: when values are close, or when there are many of them, the slices are indistinguishable and the reader leans entirely on the printed numbers (at which point you've drawn a table, not a chart). A pie earns its place only with few categories and clearly different sizes (think 60/30/10). For ranking near-equal values like these, a sorted bar chart shows the differences honestly. Choose the pie for "it's roughly two-thirds versus one-third," and a bar for everything else.

Troubleshooting