Skip to content

Instantly share code, notes, and snippets.

@chalg

chalg/u_cameco_wna_event_curve.py

Created September 20, 2025 23:21
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save chalg/a13ad329f545112ff0c37b8b1a4c7d5b to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save chalg/a13ad329f545112ff0c37b8b1a4c7d5b to your computer and use it in GitHub Desktop.
Download ZIP
Event study analysis of uranium futures and Cameco (CCO.TO) stock performance around World Nuclear Association annual symposiums (2007-2025)
Raw
u_cameco_wna_event_curve.py
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import yfinance as yf
# ---------- Reload & prep uranium data ----------
uranium_path = "data/Uranium Futures Historical Data.csv"
u = pd.read_csv(uranium_path)
u['Date'] = pd.to_datetime(u['Date'], errors='coerce')
u['Price'] = pd.to_numeric(u['Price'], errors='coerce')
u = u.dropna(subset=['Date', 'Price']).sort_values('Date').reset_index(drop=True)
u = u.set_index('Date').asfreq('B').ffill() # business days
u['ret'] = u['Price'].pct_change()
u = u.dropna(subset=['ret'])
# ---------- Download and prep CCO.TO stock data ----------
print("Downloading CCO.TO data...")
cco = yf.download('CCO.TO', start='2000-01-01', end='2025-09-21', progress=False)
cco = cco[['Close']].copy()
cco.columns = ['Price']
cco = cco.dropna()
cco = cco.asfreq('B').ffill() # business days, forward fill
cco['ret'] = cco['Price'].pct_change()
cco = cco.dropna(subset=['ret'])
print(f"Uranium data: {u.index.min()} to {u.index.max()}")
print(f"CCO.TO data: {cco.index.min()} to {cco.index.max()}")
# ---------- Helper: approximate WNA date (first Thu on/after Sep 5) ----------
import datetime as dt
def approximate_wna_middate(year: int) -> pd.Timestamp:
d = dt.date(year, 9, 5) # Sep 5
# weekday(): Mon=0 ... Sun=6 ; Thursday=3
days_to_thu = (3 - d.weekday()) % 7
d = d + dt.timedelta(days=days_to_thu)
return pd.Timestamp(d)
# ---------- Manual overrides for years you know exactly ----------
manual_event_dates = {
2015: "2015-09-10",
2016: "2016-09-15",
2017: "2017-09-14",
2018: "2018-09-06",
2019: "2019-09-05",
# 2020 omitted (pandemic uncertainty)
2021: "2021-09-09",
2022: "2022-09-08",
2023: "2023-09-07",
2024: "2024-09-05",
2025: "2025-09-04" # Added 2025 date
}
# Years to exclude explicitly (Fukushima) + anything else you want to skip
exclude_years = {2011, 2020}
# Build per-year event index using manual if available, else approximation
# Use overlapping period for both datasets
use_years = range(max(u.index.min().year, cco.index.min().year, 2000),
min(u.index.max().year, cco.index.max().year, 2025) + 1)
event_idx = {}
event_idx_2025 = {} # Separate index for 2025
for y in use_years:
if y in exclude_years:
continue # <-- skip 2011 (and 2020)
d = pd.Timestamp(manual_event_dates[y]) if y in manual_event_dates else approximate_wna_middate(y)
# Check if date exists in both datasets
if (d < u.index.min() or d > u.index.max() or
d < cco.index.min() or d > cco.index.max()):
continue
d_nearest = u.index[u.index.get_indexer([d], method='nearest')][0]
if y == 2025:
event_idx_2025[y] = d_nearest # Store 2025 separately
else:
event_idx[y] = d_nearest # Historical events for averaging
print(f"Historical events used: {len(event_idx)} years → {sorted(event_idx.keys())}")
print(f"2025 event: {sorted(event_idx_2025.keys())}")
# ---------- Function to calculate event-window cumulative returns ----------
def calculate_event_returns(data, event_idx, pre=30, post=90):
window = np.arange(-pre, post + 1)
paths = []
for y, d in event_idx.items():
# pull the window of returns
start_date = d + pd.tseries.offsets.BDay(window[0])
end_date = d + pd.tseries.offsets.BDay(window[-1])
rets = data.loc[start_date:end_date]['ret']
if len(rets) != len(window):
# incomplete window at dataset edges
continue
# cumulative path and normalize so t=0 == 1.0
cum = (1 + rets.values).cumprod()
base_idx = np.where(window == 0)[0][0]
cum_norm = cum / cum[base_idx]
paths.append(cum_norm)
paths = np.array(paths)
avg_path = paths.mean(axis=0) if len(paths) else np.array([])
std_path = paths.std(axis=0) if len(paths) else np.array([])
n_events = len(paths)
se_path = (std_path / np.sqrt(n_events)) if n_events else std_path
return window, avg_path, se_path, n_events
# ---------- Function to get single event trajectory (for 2025) ----------
def get_single_event_trajectory(data, event_date, pre=30, post=90):
start_date = event_date + pd.tseries.offsets.BDay(-pre)
end_date = event_date + pd.tseries.offsets.BDay(post)
# Get available data within the window
try:
available_data = data.loc[start_date:end_date]
if len(available_data) == 0:
return np.array([]), np.array([])
# Calculate actual window positions based on available data
actual_window = []
# Find the event date position in available data
event_pos = None
for i, date in enumerate(available_data.index):
days_diff = (date - event_date).days
actual_window.append(days_diff)
if event_pos is None and date >= event_date:
event_pos = i
if event_pos is None:
return np.array([]), np.array([])
# Calculate cumulative returns and normalize to event date
rets = available_data['ret'].values
cum = (1 + rets).cumprod()
cum_norm = cum / cum[event_pos]
return np.array(actual_window), cum_norm
except KeyError:
return np.array([]), np.array([])
# ---------- Calculate event returns for historical data ----------
pre, post = 30, 90
window_u, avg_path_u, se_path_u, n_events_u = calculate_event_returns(u, event_idx, pre, post)
window_cco, avg_path_cco, se_path_cco, n_events_cco = calculate_event_returns(cco, event_idx, pre, post)
# ---------- Calculate 2025 trajectories ----------
trajectory_2025_u = np.array([])
trajectory_2025_cco = np.array([])
window_2025_u = np.array([])
window_2025_cco = np.array([])
if 2025 in event_idx_2025:
event_2025 = event_idx_2025[2025]
window_2025_u, trajectory_2025_u = get_single_event_trajectory(u, event_2025, pre, post)
window_2025_cco, trajectory_2025_cco = get_single_event_trajectory(cco, event_2025, pre, post)
print(f"Historical Uranium futures: {n_events_u} complete event windows")
print(f"Historical CCO.TO stock: {n_events_cco} complete event windows")
print(f"2025 Uranium trajectory points: {len(trajectory_2025_u)}")
print(f"2025 CCO.TO trajectory points: {len(trajectory_2025_cco)}")
# ---------- Plot ----------
out_path = "images/uranium_event_curve_with_cco_2025.png"
fig, ax = plt.subplots(figsize=(8, 6)) #was (10, 6)
# Plot historical averages
if len(avg_path_u):
ax.plot(window_u, (avg_path_u - 1.0) * 100,
label=f'Uranium Futures Average (n={n_events_u})',
linewidth=2, color='orange', alpha=0.8)
ax.fill_between(window_u,
(avg_path_u - 1.0 - se_path_u) * 100,
(avg_path_u - 1.0 + se_path_u) * 100,
alpha=0.15, color='orange')
if len(avg_path_cco):
ax.plot(window_cco, (avg_path_cco - 1.0) * 100,
label=f'CCO.TO Average (n={n_events_cco})',
linewidth=2, color='blue', alpha=0.8)
ax.fill_between(window_cco,
(avg_path_cco - 1.0 - se_path_cco) * 100,
(avg_path_cco - 1.0 + se_path_cco) * 100,
alpha=0.15, color='blue')
# Plot 2025 trajectories
if len(trajectory_2025_u):
ax.plot(window_2025_u, (trajectory_2025_u - 1.0) * 100,
label='Uranium 2025 (Current)',
linewidth=3, color='red', linestyle='-', alpha=0.9)
if len(trajectory_2025_cco):
ax.plot(window_2025_cco, (trajectory_2025_cco - 1.0) * 100,
label='CCO.TO 2025 (Current)',
linewidth=3, color='darkgreen', linestyle='-', alpha=0.9)
ax.axvline(0, linestyle='--', linewidth=1, color='black', alpha=0.7)
ax.axhline(0, linestyle='-', linewidth=0.5, color='black', alpha=0.3)
ax.set_xlabel('Business days relative to event (0 = WNA Symposium mid-date)')
ax.set_ylabel('Cumulative return (%)')
ax.set_title('Uranium Futures vs Cameco: Historical Average vs 2025 Current Trajectory\nAround WNA Symposium')
ax.legend(loc='best')
ax.grid(True, linestyle=':', alpha=0.7)
# Add caption
caption = "Sources: Uranium - investing.com; CCO.TO - Yahoo Finance"
ax.text(
0.98, 0.02,
caption,
transform=ax.transAxes,
ha='right', va='bottom',
fontsize=8, fontstyle='italic', color='grey',
bbox=dict(facecolor='white', alpha=0.6, edgecolor='none', pad=1.5)
)
fig.tight_layout()
fig.savefig(out_path, bbox_inches='tight', dpi=300)
plt.show()
plt.close(fig)
# ---------- Print summary statistics ----------
print("\n" + "="*60)
print("SUMMARY STATISTICS")
print("="*60)
if len(avg_path_u):
final_ret_u = (avg_path_u[-1] - 1.0) * 100
print(f"Historical Uranium Futures Average - Final cumulative return (+{post} days): {final_ret_u:.2f}%")
if len(avg_path_cco):
final_ret_cco = (avg_path_cco[-1] - 1.0) * 100
print(f"Historical CCO.TO Average - Final cumulative return (+{post} days): {final_ret_cco:.2f}%")
if len(trajectory_2025_u):
current_ret_u = (trajectory_2025_u[-1] - 1.0) * 100
days_from_event = window_2025_u[-1] if len(window_2025_u) else 0
print(f"2025 Uranium Current - Return at +{days_from_event:.0f} days: {current_ret_u:.2f}%")
if len(trajectory_2025_cco):
current_ret_cco = (trajectory_2025_cco[-1] - 1.0) * 100
days_from_event = window_2025_cco[-1] if len(window_2025_cco) else 0
print(f"2025 CCO.TO Current - Return at +{days_from_event:.0f} days: {current_ret_cco:.2f}%")
print(f"\nHistorical event years: {sorted(event_idx.keys())}")
if event_idx_2025:
print(f"Current trajectory year: {sorted(event_idx_2025.keys())}")
print(f"Output saved to: {out_path}")
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment