Reference

How to tell whether a BaZi calculator is any good

Six checks you can run on any calculator in ten minutes. The point is not to find a perfect one — it is to find one whose mistakes you can see.

Updated October 2026

Why the question is harder than it looks

Enter the same birth moment into five BaZi calculators and you will often get three different charts. That is not because the underlying rules are ambiguous — the four pillars are determined by a well-defined procedure — but because the hard part is not the rules. It is everything before them: the timezone, the daylight-saving offset in force on that date, and the conversion to true solar time.

A calculator that skips those steps still produces a confident-looking chart. It is just the chart for a different moment.

So the useful question is not “which one is accurate”, which you cannot check from the outside. It is “which one shows enough for me to check it myself”. Here are six things to look for.

1. Does it ask for a city, not just a time?

The traditional reckoning runs on local solar time, not clock time. Clock time is a political artefact: it is whatever offset the relevant government had legislated on that date, and those have changed repeatedly. Solar time depends on where you actually were.

Without a birthplace, a calculator cannot correct for longitude, which shifts the reading by up to an hour across a single timezone. If it only asks for a date and a time, it is implicitly assuming you were born at the centre of your timezone. Most people were not.

The check: if it does not ask where you were born, treat the hour pillar as unknown.

2. Does it handle historical daylight saving?

This is where most calculators quietly fail. Daylight-saving rules are not a fixed formula — they vary by country, by year, and in some places by region, and several countries have changed or abolished them within living memory.

A calculator using a fixed offset, or using today's offset, will be wrong by an hour for anyone born in a summer month in a country that observed DST at the time. An hour is often enough to move the hour pillar.

The check: test it with two births a few months apart in the same city, one in winter and one in summer, in a country with daylight saving. If the time correction is identical for both, DST is not being handled.

3. Does it show the true solar time it used?

This is the single most useful thing a calculator can display, and most do not. If it shows you the clock time and the true solar time side by side, you can see the size of the correction it applied and sanity-check it against your longitude.

A correction of a few minutes is normal near the centre of a timezone. Twenty to forty minutes is normal near the edge. If the number looks wrong for where you were born, something upstream is wrong.

The check: if the page never shows the corrected time, you have no way to audit any of the steps above.

4. Does it show the pillars as characters?

A chart is four pairs of characters. If a calculator only gives you a summary — a type, a score, a colour, a personality label — then it has converted a calculation into an opinion, and you cannot check either the calculation or the opinion.

Showing the characters is not decoration. It is what makes the output comparable against any other calculator, or against a printed reference, or against a practitioner. Without them there is nothing to compare.

The check: can you take the output to a second calculator and compare character for character? If not, the first one is asking for your trust rather than your scrutiny.

5. Does it say what it does when you do not know your birth time?

Plenty of people do not know their birth hour. The honest answers are limited: either drop the hour pillar and say so, or flag the reading as incomplete. The dishonest answer is to silently assume noon, which produces a confident hour pillar that is probably wrong.

Watch specifically for whether the missing-time case changes what the page claims, or whether it just fills in a default and says nothing.

The check: run it twice, once with a time and once with the time left blank. If the output looks equally confident both times, the second one is invented.

6. Can you see it being tested?

This is the check almost nobody runs, and it is the most informative one. A calculator that publishes its test cases — a set of birth moments with the expected chart for each — is telling you what it believes the right answers are. That is a claim you can inspect.

Better still is a calculator that checks itself against a second, independently written implementation. Two implementations that share no code agreeing on the same thirty cases is real evidence. One implementation agreeing with itself is not.

The check: look for a page that shows test cases and their results. If it exists, read it. If it does not, ask why not — the test data is not difficult to publish.

What none of this tells you

These six checks are about whether the calculation is trustworthy. They say nothing about whether the interpretation is any good, which is a separate question with a separate set of standards, and one where reasonable practitioners disagree with each other.

They also do not tell you which calculator is most accurate. That is not knowable from outside, because accuracy depends on data — timezone history, city coordinates — that you cannot audit without access.

What they tell you is whether the thing in front of you is checkable. A checkable calculator with an occasional error is more useful than an opaque one that happens to be right, because only the first one can be corrected.

Applying this to WuWay

For completeness, since this site is a BaZi calculator and the question above is fair to ask of it:

• It asks for the birth city from an embedded city dataset, and uses that for both timezone and longitude.
• Offsets come from the ICU timezone database, so historical daylight-saving rules are applied as they stood on the date.
• The true solar time and the size of the total correction are shown on the result page.
• The four pillars are shown as characters.
• If the birth time is unknown, the hour pillar is dropped and the reading says so.
• Thirty reference charts with expected values are published, and the engine is checked against an independent implementation that shares no code with it. The results are at Algorithm & accuracy, and the test data itself is published as JSON.

Whether that adds up to a good reading is your call. The point of listing it is that all six checks are answerable without taking anything on trust.