A calibration curve (reliability diagram) visualizes how well the predicted probabilities of a binary classifier match actual outcomes. By plotting the fraction of positives against mean predicted probability in binned intervals, it reveals whether a model is well-calibrated, overconfident, or underconfident. A perfectly calibrated model follows the diagonal line where predicted probability equals observed frequency.
""" anyplot.ai
calibration-curve: Calibration Curve
Library: pygal 3.1.0 | Python 3.13.13
Quality: 90/100 | Updated: 2026-05-10
"""
import os
import numpy as np
import pygal
from pygal.style import Style
# Theme tokens
THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_MUTED = "#6B6A63" if THEME == "light" else "#A8A79F"
IMPRINT = ("#009E73", "#C475FD", "#4467A3", "#BD8233", "#AE3030", "#2ABCCD", "#954477")
# Data: Generate synthetic binary classification with realistic calibration
np.random.seed(42)
n_samples = 2000
n_bins = 10
# Generate true probabilities spread across 0-1 range
true_prob = np.random.beta(2, 2, n_samples)
y_true = (np.random.random(n_samples) < true_prob).astype(int)
# Model 1: Well-calibrated model (Logistic Regression style)
noise1 = np.random.randn(n_samples) * 0.08
y_prob_model1 = np.clip(true_prob + noise1, 0.01, 0.99)
# Model 2: Overconfident model (Random Forest / Neural Network style)
y_prob_model2 = 1 / (1 + np.exp(-12 * (true_prob - 0.5)))
y_prob_model2 = np.clip(y_prob_model2 + np.random.randn(n_samples) * 0.02, 0.02, 0.98)
# Compute calibration data inline
bin_edges = np.linspace(0, 1, n_bins + 1)
# Model 1 calibration
bin_indices1 = np.digitize(y_prob_model1, bin_edges[1:-1])
mean_pred1 = []
frac_pos1 = []
for i in range(n_bins):
mask = bin_indices1 == i
if mask.sum() > 0:
mean_pred1.append(np.mean(y_prob_model1[mask]))
frac_pos1.append(np.mean(y_true[mask]))
# Model 2 calibration
bin_indices2 = np.digitize(y_prob_model2, bin_edges[1:-1])
mean_pred2 = []
frac_pos2 = []
for i in range(n_bins):
mask = bin_indices2 == i
if mask.sum() > 0:
mean_pred2.append(np.mean(y_prob_model2[mask]))
frac_pos2.append(np.mean(y_true[mask]))
# Compute Brier scores
brier1 = np.mean((y_prob_model1 - y_true) ** 2)
brier2 = np.mean((y_prob_model2 - y_true) ** 2)
# Custom style with theme-adaptive colors
custom_style = Style(
background=PAGE_BG,
plot_background=PAGE_BG,
foreground=INK,
foreground_strong=INK,
foreground_subtle=INK_MUTED,
colors=IMPRINT,
title_font_size=28,
label_font_size=22,
major_label_font_size=18,
legend_font_size=16,
value_font_size=14,
stroke_width=3,
)
# Create XY chart for calibration curve
chart = pygal.XY(
style=custom_style,
width=4800,
height=2700,
title="calibration-curve · pygal · anyplot.ai",
x_title="Mean Predicted Probability",
y_title="Fraction of Positives",
show_dots=True,
dots_size=16,
stroke_style={"width": 5},
show_x_guides=True,
show_y_guides=True,
x_value_formatter=lambda x: f"{x:.1f}",
range=(0, 1),
xrange=(0, 1),
legend_at_bottom=True,
legend_at_bottom_columns=3,
legend_box_size=28,
truncate_legend=-1,
margin=50,
margin_top=80,
margin_bottom=200,
)
# Perfect calibration line (diagonal reference)
perfect_calibration = [
{"value": (0, 0), "label": "Perfect calibration reference"},
{"value": (0.25, 0.25), "label": "Predicted = Observed"},
{"value": (0.5, 0.5), "label": "Ideal: 50% predicted → 50% positive"},
{"value": (0.75, 0.75), "label": "Predicted = Observed"},
{"value": (1.0, 1.0), "label": "Perfect calibration reference"},
]
chart.add("Perfect Calibration", perfect_calibration, stroke_dasharray="15,8", dots_size=0, stroke_style={"width": 4})
# Model 1 calibration curve - well-calibrated
model1_points = [{"value": (0.0, 0.0), "label": "Curve start"}]
model1_points.extend(
[
{"value": (pred, obs), "label": f"Bin: {pred:.2f} pred → {obs:.2f} actual ({int(obs * 100)}% positive)"}
for pred, obs in zip(mean_pred1, frac_pos1, strict=False)
]
)
model1_points.append({"value": (1.0, 1.0), "label": "Curve end"})
chart.add(f"Logistic Regression (Brier: {brier1:.3f})", model1_points)
# Model 2 calibration curve - overconfident
model2_points = [{"value": (0.0, 0.0), "label": "Curve start"}]
model2_points.extend(
[
{"value": (pred, obs), "label": f"Bin: {pred:.2f} pred → {obs:.2f} actual (Δ={pred - obs:+.2f})"}
for pred, obs in zip(mean_pred2, frac_pos2, strict=False)
]
)
model2_points.append({"value": (1.0, 1.0), "label": "Curve end"})
chart.add(f"Overconfident Model (Brier: {brier2:.3f})", model2_points)
# Save output files
output_dir = os.path.dirname(os.path.abspath(__file__))
chart.render_to_png(os.path.join(output_dir, f"plot-{THEME}.png"))
chart.render_to_file(os.path.join(output_dir, f"plot-{THEME}.html"))
Part of Calibration Curve on anyplot.ai.