BaZi vs Munger’s Latticework: One Step Short

The point of statistics is not observation. It is calibration. It demands three things: a measurable outcome, a prediction stated in advance, and a test that is allowed to fail. Without the third, any amount of accumulated observation is still a classification, not a statistic.

That the BaZi system is a product of long observation is not in dispute. What it never set out to do is those three things.

What it hands back is not a probability. It is a position: which part of your chart carries the weight, which kind of reaction keeps recurring, under what conditions the balance turns. A position can be described in fine detail — possibly finer than you have ever described yourself. It does not come with a percentage.

In 1994 Charlie Munger gave a talk at the University of Southern California’s Marshall School of Business titled “A Lesson on Elementary, Worldly Wisdom,” later collected in Poor Charlie’s Almanack. What he described is the same problem approached the other way round.

The word he used was not knowledge. It was a latticework of mental models. The metaphor is doing real work: not a list, not a toolbox, but a structure whose pieces hold each other up — a new fact has to hang somewhere on it, and be checkable against several rungs at once. In his words, if you only remember isolated facts you do not have them in a usable form; facts have to hang together on a latticework of theory before you can be said to own them.

How the latticework runs he spelled out himself. It comes apart into four steps.

One: the models are borrowed, not invented. They come from physics, biology, mathematics, psychology, engineering, economics — not from his own trade. He was explicit that other fields had spent centuries grinding them out through experiment and error; his contribution was laying them out next to each other.

Two: there are few of them. His estimate was that eighty or ninety important models carry about ninety per cent of the freight — not thousands, but a set small enough to actually finish learning.

“80 or 90 important models will carry about 90% of the freight in making you a worldly-wise person.” Charlie Munger, “A Lesson on Elementary, Worldly Wisdom,” USC Marshall School of Business, 1994, collected in Poor Charlie’s Almanack

Three: several have to be running at once. His own warning was that if you use only one or two, the nature of human psychology is such that you will torture reality until it fits the model. A single model is not merely insufficient. It actively edits what you are able to see.

Four: the power is in the pile-up. He gave this a name — the lollapalooza effect. When several models push in the same direction at once, the result is not additive; it blows out. And you can only see it coming if you are holding more than one model at a time.

There was one more move he kept coming back to: invert. He borrowed it from the mathematician Carl Jacobi — “invert, always invert.” Before you argue that a plan will succeed, argue how it would fail. He also built his own catalogue of the psychology of human misjudgement — twenty-five cognitive tendencies, the rung he reached for most often.

Put the two side by side and the outline is clear. BaZi draws on the five phases, the ten gods, the palaces, the four transformations. Munger draws on compounding, incentives, redundancy, ecological niches, the psychology of misjudgement. Both do the same thing: compress long observation into a small set of reusable categories. As raw material, they are the same kind of thing.

The difference is in the next section, and it is far more concrete than “accuracy.”

To take “BaZi is observed and generalised” apart properly you have to look at where its parts came from. There is a stretch of history that usually gets skipped.

The ten stems and twelve branches were not invented to read fortunes. They were invented to count days. On excavated Shang oracle bones the date of a divination is written in stems and branches — divined on bing-yin, divined on gui-you — and one ox scapula carries the complete sixty-term cycle. Shang kings named themselves after stems: Tai Yi, Pan Geng, Wu Ding. The original job of the stems and branches was the calendar, an official timekeeping tool, with nothing to do with a person’s character or fate.

And half of where they came from is genuine astronomical observation. In the Spring and Autumn period there was a system called year-star reckoning: the ancients noticed Jupiter takes about twelve years to come round the sky, divided the ecliptic into twelve regions, and used it to count years. When they later found the real period to be a little over eleven years and eight months, they invented a notional counter-star, Tai Sui, to absorb the drift — exactly twelve years to the round. Only in the Eastern Han did the stems-and-branches year count still in use today settle into its final form.

Here is the distinction that matters: what was observed was the sky, not people. What accumulated was calendar precision. Not one person was sampled in that record. Not one person’s fortunes were logged for comparison.

Fate reading is something else, and much later. In the Tang dynasty Li Xuzhong began reading human affairs off the year, month and day of birth — the three pillars. In the Song, Xu Ziping added the hour pillar; by the late Southern Song the method was fixed on the five phases rather than the sound-elements, and the four-pillar chart took shape — more than a thousand years after the stems and branches had first been used to keep a calendar.

So “BaZi is a product of long observation” folds two things into one. The timekeeping half does come from long observation of the sky, and that need not be denied. But mapping the marks on a calendar onto a person’s temperament and fortunes is analogical reasoning, not observation and statistics. There is a whole link missing in the middle: no people.

The observation is real. It was observation of the sky. Reading the sky’s cycles as the regularities of human life is a further step, and that step has no sample.

Each of Munger’s eighty-odd models carries external verification. Get compounding wrong and the money simply is not there. Skimp on redundancy and the bridge comes down. Set the incentive backwards and people act in ways you did not predict. Those models survived because they were falsified repeatedly — the wrong parts were cut away long ago.

The BaZi system never goes through that process. It has no sample, no control group, and no mechanism that retires a rule that fails. Its tolerance runs the other way: so high it is almost boundless. When a reading does not come true there is always one level further up to check — the wrong major cycle, the wrong useful god, the wrong birth hour. When it does come true, that is kept as evidence. Nobody is lying here. The system simply has no built-in way to be contradicted.

Which is why “that feels accurate” is not a test. It has another name: hindsight attribution. After the fact, we can construct a reason why almost any sentence applied to us — and that says nothing about the sentence.

Munger quoted a proverb for this — to the man with only a hammer, every problem looks like a nail. His own phrasing is sharper: use only one or two models and you will torture reality until it fits. One model distorts judgement; that is exactly why he wanted a latticework. The interesting part is that the BaZi system is multi-dimensional by construction: BaZi reads the distribution of resources and the tempo, Zi Wei reads the palaces and the relationships, and where the two charts disagree is where the real question lives.

The ruler has three marks: can it be quantified, can a prediction be stated in advance, can it be contradicted. Now hold it against the two things most often set beside BaZi — astrology and the Yijing. We are only laying out structure here. No score.

Astrology. Its observational half is very hard. Mesopotamia was logging the sky from around 2000 BCE — eclipses, planetary positions — and eventually compiled the omen series Enuma Anu Enlil, dozens of tablets in the form: this celestial event was observed, therefore this will follow. A twelve-sign zodiac appears by the fifth century BCE, and the oldest surviving individual birth chart is dated 410 BCE. After Alexander, the Babylonian tables fused with Greek natural philosophy, and Ptolemy wrote the Tetrabiblos in the second century CE, trying to give it a causal footing — the sun and moon acting on climate and biological cycles. That step points toward something testable.

But the bridge from sky to human affairs is a correspondence, not a correlation. So the first two marks are answered loosely. The third was put to it directly in history. Around 156 BCE the Greek sceptic Carneades raised a plain objection: people born at the same moment meet wildly different fates, while people born at different moments die together in the same shipwreck. Cicero wrote up arguments of this kind in On Divination, and Augustine later pressed it with the case of twins. That is the ancients holding it against the sample themselves, not a charge filed by posterity.

The Yijing. Its starting point is clearer still — divination. Shang divination was pervasive and oracle bones carry thousands of records; on Shang and Zhou bronzes and shells there is also a number hexagram, six values drawn from six, seven, eight and nine recording the outcome of one casting, later simplified into yin and yang lines. The Yijing’s first identity was an official divination manual, alongside the Lianshan and the Guicang. Its final text was fixed around the end of Yin and the start of Zhou, and modern scholars mostly take it as compiled by several hands over time, not written by one sage king.

What is interesting is the road it took afterwards. By the Spring and Autumn and Warring States periods, Confucian scholars added the Ten Wings — lines such as “one yin and one yang, that is the Way” — and a divination handbook was re-read as a philosophy of change. So Yi has worn two faces historically. As a tool for getting answers it runs on omens; as a frame for watching change it runs on structure. Which face you are looking at, the ancients had already sorted out once themselves.

Three systems, laid out here, measured with one ruler. Where each of them holds and where it does not — your answer will not be identical to ours, and that is as it should be. Ours is to lay the structure out. Judgement is yours.

The honest label is a language for describing structure, not a forecasting tool.

The job of a language is not prediction. It is naming. Most people hold their own state as a fog — “it has been off lately,” “why does this keep happening again.” A classification system cuts the fog into parts: not enough resource, wrong tempo, or the same reaction at the same position every time. You can only work on what you can name.

This is the part it genuinely shares with Munger’s models. Neither decides for you. Both swap in a different lens so you look at the same situation again. His latticework answers “you are holding one hammer”; reading two charts answers “you are looking from one angle.” Neither is an answer. Both are tools that let you ask a better question.

Whether it is usable is not settled by “is it accurate.” It is settled by whether it made you ask something you could not have asked before.

One: is what I was given “which kind of reaction you lean toward”, or “what will happen to you this year”? The first can be checked against yourself again and again. The second can only be awaited.

Two: did I ever write it down before the event rather than after it? Saying “I knew it” afterwards does not count as a test.

Three: the last time it was wrong, what did I actually do with it? If the answer is “something must have been calculated wrong,” then that filtering step does not exist.

Four: if I put the same three questions to astrology or the Yijing, would I accept the same kind of answer? If I am strict with one and lenient with another, then what I am holding is not a ruler. It is a position.

Is BaZi actually a form of statistics?

No. The observational side is real, but the core of statistics is calibration: measurable, stated in advance, allowed to fail. BaZi is missing the third. It has no built-in mechanism to be contradicted, which makes it a system of classification and description rather than a statistical model. We use it only in that descriptive sense.

Then is it the same thing as Munger’s mental models?

Same material, different process. Both compress long observation into reusable categories. But each of Munger’s models is tied to external verification — the arithmetic is wrong, the bridge falls, people respond to the incentive backwards. BaZi never goes through that step. Not the same tool; the same material through a different process.

Does “that feels accurate” count as evidence?

No. After an event we can construct a reason why almost any sentence applied to us. That is hindsight attribution. To count as evidence it has to be written down before the event and compared afterwards — which is why our reports keep conclusions that are available separate from conclusions limited by the precision of your birth time.

How does Baize handle this?

No scores, and no pronouncements of good or bad fortune. We cross-check two charts; when BaZi and Zi Wei point in different directions we publish both, including where they disagree. The reports describe structure and a next step that can be acted on. They do not promise outcomes.

Are astrology and the Yijing statistics, then?

We are not going to hand you a verdict here. Hold them against the same ruler: can it be quantified, can a prediction be stated in advance, can it be contradicted. Astrology’s record of the sky is older and finer than anyone else’s, but the bridge from sky to human affairs is a correspondence; the Yijing’s first job was divination, and it was later re-read as a philosophy of change. Which mark each of them holds is your call to make. For what it is worth, someone was already putting the sample to astrology in the second century BCE — a Greek sceptic, not a modern.

The stems and branches did come from astronomical observation, though?

What they observed was the sky, not people. The stems and branches began as a calendar for counting days, and the twelve branches do trace back to real observation of Jupiter’s cycle. But no person was sampled in that record, and no one’s fortunes were logged for comparison. Mapping the marks on a calendar onto a person’s temperament and fortunes is a step of analogy, not observation.

This piece is not about whether Chinese metaphysics works. It is about what you are using it for. If the language is useful to you it should make you ask something you could not have asked before. These lines are safe to copy:

Ask Baize these first — copy them as they are:

• “Which positions on my chart are the ones that keep walking me back to the same road?”
• “When I judge a situation, which category of information do I systematically leave out?”

It answers against your chart, not as a generic assistant.

To see the “same road coming back” laid out properly, use the Decision Blind Spots & Improvable Fixed Patterns report: how you default to deciding, which category of information you repeatedly miss, and which pattern is fastest to move. It gives description and an executable next step — not a promised outcome.

First time on Baize and unsure where to start? The full order, from reading the chart to asking and reviewing, is in “Opening Baize for the first time: where to begin.”

This page is written for cultural study and self-reflection. It makes no predictive promise and is not medical, legal, investment or career advice.