Stacked Bar Charts: Win-Loss Composition

Here's the catch nobody mentions: a stacked bar chart is great until it quietly lies to you, and most tutorials never show you the difference. A plain bar compares totals; a stacked bar shows what those totals are made of — stack a team's wins and losses into one bar and the bar's length is games played, while the split inside shows the record. We'll build a win-loss stacked bar for the best teams of 2023, then pin down the one situation where stacking genuinely helps and the one where it misleads.

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

1. The data: two parts of one whole

   Each team's season splits cleanly into wins and losses that sum to games played — the natural shape for a stacked bar.

   import pandas as pd
   
   df = pd.read_csv("sample_standings.csv")
   print(df.sort_values("W", ascending=False)[["Team", "W", "L", "G"]].head(6).to_string())

   Wins, losses, and games

         Team    W   L    G
   0   Braves  104  58  162
   1  Orioles  101  61  162
   2  Dodgers  100  62  162
   3     Rays   99  63  162
   4  Brewers   92  70  162
   5   Astros   90  72  162

   Every team played 162 games, so the Braves' 104–58 and the Dodgers' 100–62 will produce bars of identical length — what differs is where the wins-to-losses line falls inside each.

2. Stack with the left argument

   Stacking is just drawing the second series starting where the first ends. For a horizontal chart, that means giving the losses a left= equal to the wins, so they begin at the end of the win segment.

   import matplotlib.pyplot as plt
   
   top = df.sort_values("W", ascending=False).head(10).sort_values("W")
   fig, ax = plt.subplots(figsize=(8, 6))
   ax.barh(top["Abbr"], top["W"], label="wins")
   ax.barh(top["Abbr"], top["L"], left=top["W"], label="losses")   # stack starts at W
   ax.set_xlabel("games"); ax.legend(loc="lower right")
   fig.savefig("win_loss_stacked.png", dpi=144, bbox_inches="tight")

   

   For a vertical bar chart the same idea uses bottom= instead of left=. Either way, the rule is: each layer starts where the stack so far ends. Here every bar reaches 162, and the eye reads each team's record off where the color changes.

3. When stacking helps — and when it lies

   Stacking shines when the parts genuinely sum to a meaningful whole and you care about composition: win-loss out of games, or scoring broken into singles/doubles/etc. It works because the first segment shares a common baseline (zero), so those segments are easy to compare. The trap is the upper segments: because losses all start at different points, comparing loss totals by eye is hard — your eye has to subtract. If the losses were the comparison you cared about, a grouped bar (or just plotting losses alone) would be honest where a stack is not. Choose the chart for the comparison you want the reader to make.

Troubleshooting