What BaZi Cannot Tell You

The chart this blog has used from the start was cast for 04:00 on 28 February 1990, in Seoul: 庚午, 戊寅, 甲子, 丙寅. I've used it to show how the luck pillars run and what a Fire Horse year brings. There's one thing I haven't said about it. It isn't one person's chart.

Most of what this blog calls accuracy is about getting a chart right: the solar terms to within about ten minutes, the hour corrected for longitude. This post is about the question that comes after, the one the last post ended on. Once the eight characters are right, does anything they say hold up when it's tested?

A chart is a two-hour window

A chart isn't cast for a person. It's cast for a moment, and in BaZi a moment is two hours wide. The hour pillar is the finest of the four, and it changes every two hours.

Everyone born in the 寅 hour on 28 February 1990, from three to five in the morning by local solar time, has the example chart. That's true in Seoul and anywhere else, because the year and the month were the same everywhere that day, and the day pillar follows the local date. A baby born at four that morning in New York, fourteen hours after the one in Seoul, has the same eight characters.

About 143 million babies were born worldwide in 1990, roughly 390,000 a day, or about 33,000 in an average two hours. So some 33,000 people share all eight characters of the example chart, and with them everything a reading could draw from those characters.

The arithmetic holds for any chart. There are 518,400 possible ones: sixty year pillars with twelve months each, sixty day pillars with twelve hours each. There are more than eight billion people. No chart is anyone's alone.

And charts come back. The example exists for two hours on 28 February 1990, and next for two hours on 13 February 2050, the next Jia Zi 甲子 day to fall in the 戊寅 month of a 庚午 year. Between 1900 and 2100 those are the only two mornings it exists. It skipped 1930, though 1930 was a 庚午 year too. That year's 寅 month ran from 4 February to 6 March, and no 甲子 day fell inside it; the nearest came on 15 March, nine days too late.

What a crowd can't be told

Whatever a chart says, it says to everyone born in its two hours. So anything that happens to one of them and not to the rest can't be in it: who they marry, or whether; what they do for a living; what they earn; what makes them ill; when they die. A chart that says you'll marry in 2029 says it to 33,000 people.

The timing layers barely break the crowd up. The luck pillars split it in two, by birth sex. The boys born that morning all step into 壬午 at about thirty-two; the girls all step into 乙亥 at about twenty-eight. Where they were born moves the handover by a few months at most: the New York boy's first luck pillar starts at one year ten months, the Seoul boy's at two years. And the year pillar, as the last post said, is the same for everyone alive.

The crowd of 2050 will get the same eight characters, the same luck pillars in the same order, and the same years turning over them, sixty years later. Only the handover ages move, because 13 February sits at a different distance from the solar terms than 28 February does. Those boys will reach 壬午 at about thirty-seven, and the girls 乙亥 at about twenty-three.

So here are two crowds sixty years apart, and BaZi reads them as nearly the same life. Everything that will make those lives different is outside the chart: the century, the parents, the country, the work there is to do.

Born minutes apart

That much is arithmetic. It shows that a chart can't tell apart the people who share it. It doesn't show whether the chart says anything true about what they have in common. That part has been tested.

The cleanest test used people who, in BaZi's terms, had the same chart. In 2003 the researchers Geoffrey Dean and Ivan Kelly analysed data on 2,101 people born in London between 3 and 9 March 1958, collected for a study that had nothing to do with astrology. They put everyone in order of birth, which made 2,100 pairs of "time twins", born on average 4.8 minutes apart. Each person had 110 variables on record from measurements at ages 11, 16 and 23: IQ, reading and arithmetic scores; teachers' and parents' ratings of things like anxiety, aggressiveness and sociability; height, weight, vision and hearing.

In BaZi terms nearly every pair was a single chart, since a pillar that lasts two hours rarely changes inside five minutes. If the moment of birth shaped any of those 110 things, the twins should have resembled each other. They didn't. The effect attributable to astrology came out at 0.00, give or take 0.03. The test conditions, Dean and Kelly wrote, "could hardly have been more conducive to success."

The soft claim, tested

An earlier post set out the softer claim, the one this site works from: a chart describes a temperament, a lean, not events. That has been tested too.

In 1985 Nature published a double-blind test of astrology that Shawn Carlson had run as a physics student at Berkeley. The National Council for Geocosmic Research, an American astrological organisation, advised on the design and nominated the astrologers, who agreed beforehand that it was a fair test. Twenty-eight astrologers took part. For each natal chart they were given three personality profiles from a standard psychological questionnaire, the California Psychological Inventory, one of which belonged to the chart's owner. They picked the right one 40 times out of 116. Chance alone would pick it about 39 times. The astrologers advising the study had predicted at least half.

The volunteers did no better the other way round. Shown three chart readings and asked which had been written for them, 28 out of 83 picked their own. Chance predicts about 28. A 2009 reanalysis argued that the astrologers' ratings, a second measure taken in the same task, show a faint effect. Even if it's right, the first choices still came out at chance, and the astrologers had expected half.

That was Western astrology. BaZi's own version of the test is harder to find. The closest I've come across is a 2015 study in the Yonsei Medical Journal, in which two Korean researchers worked out the four pillars of 148 adults and set them against two standard personality tests. The chart's main personality markers showed no significant relationship with the scores: not the count of yin and yang across the eight characters, not whether the day master is yin or yang, not the five elements. Some correlations turned up only when the participants were regrouped by a composite yin–yang variable. It's one small study, and a result that appears only in a composite grouping, when the main markers show nothing, needs repeating before it carries much weight.

The five elements in a medical journal

The most striking result the five elements have had came from a study of how people die.

In 1993 The Lancet published a study by David Phillips and two colleagues of California death records from 1969 to 1990. It compared the deaths of 28,169 Chinese Americans with those of more than 400,000 matched white Americans. In Chinese astrology and medicine a birth year carries an element, and each element is linked to organs and kinds of illness: Metal, for example, to the lungs. Among Chinese Americans who died of bronchitis, emphysema or asthma, those born in a Metal year died at 66.9 on average, and those born in other years at 71.9. Among white Americans the same comparison came to less than a month. The pattern held for nearly all the major causes of death the study covered, and the more strongly a group held to Chinese tradition, the more years it lost.

The authors didn't conclude that the year did it. They put the effect down, at least partly, to psychosomatic processes: to belief, not to the year.

Then it was checked. In 2006 the economist Gary Smith went back over the study with a more complete statistical analysis and with independent California death records from 1960 to 1968 and from 1991 to 2002. The original results didn't replicate. The five elements' best evidence turned out to be evidence about belief, and on a second look, not even that.

Where a chart does change lives

Belief does show up elsewhere, and more solidly.

Many Chinese parents hope for a Dragon baby, and births jump when the year comes round. In a study published in 2020, the economists Naci Mocan and Han Yu looked at how Dragon children do at school, using three sets of data from China. Those born in a Dragon year were more likely to go to university and scored higher on the university entrance exam, and in middle school their test scores were higher too. On cognitive tests, the middle-schoolers were no different from other children.

The difference was their parents. Parents of Dragon children expected more of them and invested more in them, in time and money, and once the authors took those expectations into account, the Dragon effect on test scores disappeared. It even reached the children's height. Girls in the data were about six centimetres shorter than boys, and for girls born in a Dragon year that gap was cut by about half, which the authors suggest may reflect better nutrition. Neither the children nor their families were different to begin with. In the authors' words, the belief and the investment that followed it "become self-fulfilling."

The last post had the same mechanism running the other way. Japan's births fell by a quarter in 1966, and Korea's sex ratio at birth peaked in 1990, a white-horse year. A year pillar did nothing to those children. What reached them was what the adults around them believed.

It's the warning in Yuan Huang's story, too. Told his whole life in advance, he stopped wanting anything. A forecast doesn't need to be right to change a life. It only needs to be believed.

Where this site stands

So here is the list. A BaZi chart cannot tell you who you'll marry, what you'll earn, what will make you ill, or when you'll die. It cannot tell you what will happen in a decade, or a year, or on any date. It cannot tell you anything that would be different for the other people born in your two hours. And in the tests above, it showed no reliable sign of describing character either.

The readings here are written to that limit. The introductory post says a chart does not predict events. The rule the readings are written under is narrower still: a decade can ask something of you, but a reading never says what you'll do in it, and never puts an age inside it.

What's left is what the about page calls a vocabulary. BaZi is an old, carefully built way of asking questions about a life: what feeds you, what drains you, what season you were born into, which way you lean. They're good questions. The answers have to come from you, because the chart gave the same ones to everyone born in your two hours.

Your chart is what you have in common with tens of thousands of strangers. What BaZi cannot tell you is everything else.