If you have used two Bazi apps and got different Day Masters, one of them is wrong. This is not a matter of school or interpretation. The day count is arithmetic, and arithmetic has one right answer.
The sixty stem-branch combinations cycle continuously through days, without regard to months, years or calendars. Day one is followed by day two, and after sixty the cycle repeats. It has run unbroken for a very long time.
To compute the pillar for any date, you need two things: the number of days between that date and a known reference, and the stem-branch value of that reference. The arithmetic is a modulo operation and there is nothing to interpret.
This means the entire correctness of the Day Pillar rests on the reference date being right. Nothing else in the calculation can compensate for a bad anchor.
A commonly circulated reference pairs 1984-02-02 with the first combination of the cycle. It appears in tutorials, in code samples, and consequently inside a number of shipped apps. It is off by two days.
Because the error is a constant offset, it never looks like a bug. Every chart the tool produces is internally consistent and plausible. You get a Day Master, it has a full description, the reading hangs together. It is simply the wrong person's reading.
A constant offset is the most dangerous kind of error, because nothing in the output looks broken.
This calculator uses 1949-10-01 as the reference for the first combination of the cycle, cross-checked against an astronomical ephemeris and official almanac data across more than a thousand dates. Two independent verification sources agreeing on a thousand dates is a much stronger claim than any single tutorial.
You can check this yourself without trusting anyone. Take a date whose pillar you can verify in a printed almanac, then compare what your app produces. If it disagrees by exactly two days consistently, you have found the offset.
The Day Master is the reference point for the entire reading. Every Ten God label is computed relative to it. Strength is measured as support for it. The structure, the Useful God, the spouse position and most of the stars are all derived from it or from the day branch.
So a two-day error does not slightly degrade the reading. It replaces it. If your Day Master should be Yin Metal and the app says Yang Water, you have been reading advice written for a different chart, possibly for years.
Being precise about the Day Pillar makes it worth being honest about the rest. Year and month pillars here use average solar-term dates rather than computed ones, so a birth on a boundary day, roughly 3 to 5 February or the 4th to 8th of any month, can land one pillar off. Those cases are worth confirming against a precise almanac.
The hour correction uses longitude offset but not the equation of time, a seasonal variation of up to about sixteen minutes. If your birth time sits within a quarter hour of a two-hour boundary, treat the Hour Pillar as uncertain.
Stating limits like these is not a weakness in a tool. A tool that claims precision everywhere is telling you it has not checked.
Build your own chart free No signup. Everything is calculated in your browser.