The year boundary in Saju/BaZi isn't Jan 1 — it's the solar term Ipchun (立春), around Feb 4. Precisely the moment the Sun's apparent ecliptic longitude reaches 315°.

Most online 'free' Saju readers approximate this as 'Feb 4 00:00' — wrong by up to 24 hours, which flips year pillars for anyone in the window.

Correct way: Meeus astronomical algorithm — compute apparent solar longitude from Julian Day, iterate to find exact crossing.

Built one that does this: https://tarofortune.pythonanywhere.com/en