Data Visualization

A table of a thousand prices tells you almost nothing at a glance; a single chart of them tells you the whole story in a second. Visualization is where data analysis lands - where a trend, a crash, or a fat tail becomes instantly obvious. This is the capstone of Module 4, and fittingly it's wall-to-wall charts. You've already seen plots appear throughout; now we'll understand how they're built, and learn the handful of moves that cover almost everything you'll want to draw.

matplotlib: figure and axes

Every chart in this course is drawn by matplotlib (with seaborn making it pretty). The one pattern to internalise is fig, ax = plt.subplots(): it gives you a Figure (the whole canvas) and an Axes (the plot area you draw on), and you build the chart by calling methods on ax:

Copy

from pathlib import Path

import matplotlib

matplotlib.use("Agg")
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

df = pd.read_csv("reliance_6mo.csv", index_col="Date", parse_dates=True)

sns.set_theme(style="whitegrid")             # seaborn styles the whole figure
fig, ax = plt.subplots(figsize=(8, 4))       # one Figure, one Axes (ax)
ax.plot(df.index, df["Close"], color="#7c83ff", lw=1.8, label="Close")
ax.set_title("RELIANCE - daily close, last 6 months")
ax.set_xlabel("Date")
ax.set_ylabel("Price (Rs)")
ax.legend()

out = Path(__file__).with_suffix(".png")
plt.savefig(out, dpi=110, bbox_inches="tight")
print("Saved", out.name, "-", len(df), "points")

Live output

Saved 01_price_line.png - 124 points

ax.plot(x, y) draws the line; ax.set_title, ax.set_xlabel, ax.set_ylabel label it; ax.legend() adds the key; and plt.savefig(...) writes it to a file. That's the skeleton of every chart you'll build.

The shape of returns: histograms

A line shows price over time; a histogram shows the distribution - how often moves of each size happen. It's the single most revealing chart in finance, because it exposes risk that a price line hides:

Copy

from pathlib import Path

import matplotlib

matplotlib.use("Agg")
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

df = pd.read_csv("reliance_6mo.csv", index_col="Date", parse_dates=True)
ret = (df["Close"].pct_change() * 100).dropna()

# A histogram shows the SHAPE of the returns - how often each size of move happens.
sns.set_theme(style="whitegrid")
fig, ax = plt.subplots(figsize=(8, 4))
sns.histplot(ret, bins=25, kde=True, color="#7c83ff", ax=ax)
ax.axvline(0, color="#888", ls="--", lw=1)
ax.set_title("RELIANCE - distribution of daily returns")
ax.set_xlabel("Daily return (%)")

out = Path(__file__).with_suffix(".png")
plt.savefig(out, dpi=110, bbox_inches="tight")
print("Mean:", round(ret.mean(), 3), "%  Std:", round(ret.std(), 3), "%")

Live output

Mean: -0.124 %  Std: 1.499 %

The bars pile up near zero (most days are quiet) and thin out toward the edges (big moves are rare) - but notice the bars way out at -4%: those "fat tails" are exactly the dangerous days a price chart glosses over. Here we used sns.histplot with kde=True, which adds a smooth curve over the bars.

Comparing fairly: rebasing

To compare two stocks at different prices, plot them on the same chart - but first rebase each to start at 100, so you're comparing growth, not absolute rupees:

Copy

from pathlib import Path

import matplotlib

matplotlib.use("Agg")
import matplotlib.pyplot as plt
import seaborn as sns
import yfinance as yf

sns.set_theme(style="whitegrid")
fig, ax = plt.subplots(figsize=(8, 4))

# Rebase each stock to 100 so they start together - a fair growth comparison.
for ticker, color in [("AAPL", "#7c83ff"), ("MSFT", "#21c87a")]:
    close = yf.Ticker(ticker).history(period="6mo")["Close"]
    rebased = close / close.iloc[0] * 100
    ax.plot(rebased.index, rebased, lw=1.8, label=ticker, color=color)

ax.axhline(100, color="#888", ls="--", lw=1)
ax.set_title("Growth of 100 - AAPL vs MSFT, last 6 months")
ax.set_ylabel("Rebased to 100")
ax.legend()

out = Path(__file__).with_suffix(".png")
plt.savefig(out, dpi=110, bbox_inches="tight")
print("Compared 2 stocks, each rebased to 100")

Live output

Compared 2 stocks, each rebased to 100

Calling ax.plot twice (once per stock) draws both lines on one Axes, and ax.legend() labels them. Dividing each series by its first value times 100 puts both at the same starting line - now the chart answers "which grew more?" cleanly, regardless of whether one trades at 290 and the other at 1300.

Many views at once: subplots

plt.subplots(1, 2) gives you two Axes side by side in one figure - perfect for showing price and its return distribution together:

Copy

from pathlib import Path

import matplotlib

matplotlib.use("Agg")
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

df = pd.read_csv("reliance_6mo.csv", index_col="Date", parse_dates=True)
ret = (df["Close"].pct_change() * 100).dropna()

# One figure can hold several Axes side by side - here, two views at once.
sns.set_theme(style="whitegrid")
fig, (ax_left, ax_right) = plt.subplots(1, 2, figsize=(10, 4))

ax_left.plot(df.index, df["Close"], color="#7c83ff", lw=1.6)
ax_left.set_title("Price over time")
ax_left.set_ylabel("Rs")

ax_right.hist(ret, bins=20, color="#21c87a", alpha=0.85)
ax_right.set_title("Spread of daily returns")
ax_right.set_xlabel("%")

fig.suptitle("RELIANCE - two views of the same data", fontsize=13)
out = Path(__file__).with_suffix(".png")
plt.savefig(out, dpi=110, bbox_inches="tight")
print("Saved a two-panel figure")

Live output

Saved a two-panel figure

fig, (ax_left, ax_right) = plt.subplots(1, 2) hands back both Axes to draw on independently, and fig.suptitle adds one title over the whole figure. Grids of subplots are how dashboards and tear-sheets are built.

A few styling habits

Charts that communicate share a few simple habits, all of which you've now seen:

• Always add a title and axis labels - an unlabelled chart is a puzzle.
• Add a legend the moment there's more than one line (ax.legend()).
• Save crisply with dpi=110 and bbox_inches="tight" so nothing is clipped.
• Set the theme once with sns.set_theme(style="whitegrid") for a clean, consistent look.

Try it yourself

• Re-make the price line in a different colour, and add a horizontal line at the average close with ax.axhline(df["Close"].mean(), ls="--").
• Change the histogram to bins=50. Does the shape get clearer or noisier? Why might too many bins mislead?
• Add a third stock (say "GOOGL") to the rebased comparison by extending the loop's list.

Recap

• Every chart starts with fig, ax = plt.subplots() - a Figure (canvas) and an Axes (ax) you draw on.
• ax.plot for lines, ax.bar for bars, hist/sns.histplot for distributions; label with set_title/set_xlabel/set_ylabel and legend().
• seaborn styles and simplifies matplotlib; rebasing to 100 compares growth fairly; subplots put several charts in one figure.
• Save with savefig(file, dpi=110, bbox_inches="tight") - and always label your axes.

That completes Module 4 - you can now load, clean, reshape, summarise and visualise real market data. You've quietly become dangerous with data. In the final module, Python for the Markets, we turn all of it on the markets themselves: the exchanges and symbols, OHLCV bars, live quotes over HTTP, and a closing look at returns, signals and performance. It begins with a map of the market landscape.