Skip to content

Data Sources

pykabu-calendar aggregates earnings calendar data from multiple sources, each with different characteristics.

Available Sources

Source Browser Required Speed Notes
sbi Yes (Playwright) Slower Most comprehensive, default
matsui No Fast Lightweight, default
tradersweb No Fast Lightweight, default

SBI Securities

The SBI calendar is the most comprehensive public source for Japanese earnings dates.

URL: https://www.sbisec.co.jp/...

Data provided:

  • Announcement date and time
  • Company code and name

Requires: Playwright browser automation

pip install playwright
playwright install chromium

# SBI is included by default
df = cal.get_calendar("2026-02-10")

# Or use SBI only
df = cal.get_calendar("2026-02-10", sources=["sbi"])

Matsui Securities

Lightweight source that doesn't require browser automation.

URL: https://finance.matsui.co.jp/...

Data provided:

  • Announcement date and time
  • Company code and name
# Used by default
df = cal.get_calendar("2026-02-10")

# Or directly
df = cal.get_matsui("2026-02-10")

Tradersweb

Independent source that can catch discrepancies with Matsui.

URL: https://www.traders.co.jp/...

Data provided:

  • Announcement date and time
  • Company code and name
# Used by default
df = cal.get_calendar("2026-02-10")

# Or directly
df = cal.get_tradersweb("2026-02-10")

Historical Inference

Uses past announcement patterns from pykabutan to predict timing.

Logic:

  1. Fetch past N earnings announcement times
  2. Detect patterns (e.g., "always 13:00", "always after close")
  3. Assign confidence based on consistency

Example patterns:

  • Company always announces at 13:00 → high confidence
  • Company varies between 11:00-15:00 → low confidence
  • Company always announces after 15:30 → not significant for trading
# Enabled by default
df = cal.get_calendar("2026-02-10", infer_from_history=True)

# View past earnings times for a specific stock
past = cal.get_past_earnings("7203")  # Toyota

Datetime Selection Priority

When multiple sources provide different times:

  1. Inferred + Source match - When inferred time matches a source (high confidence)
  2. Inferred - From historical patterns
  3. SBI - Primary public source
  4. Matsui - Lightweight source
  5. Tradersweb - Lightweight source

The candidate_datetimes column contains all possible times, ordered by confidence.