I Counted How Many People Share the Same Zi Wei Dou Shu Chart, at Two Different Resolutions

Do two people born at the same hour end up with the same fate. The Comprehensive Book of Zi Wei Dou Shu, a Ming-dynasty compendium, already flagged the question. It puts it plainly, people born in the same year, month, day and hour still end up rich or poor, long-lived or short-lived. Whoever wrote that line had noticed the pattern, but nobody ever attached a number to it.

Last month I ran a census across 262,980 charts to measure how common each configuration is. This time the question is different. Instead of asking how common a single feature is, I asked how many distinct charts exist at all, and how many birth moments collapse onto the same one. Same scope, every day from 1 January 1965 to 31 December 2024, twelve double-hours a day, 262,980 moments total, San He transformation table, no true solar time correction, gender fixed to male. I ran the same batch of moments twice. Once counting only the Life Palace branch, the Five Element Class, and where the major stars land. Once adding the Body Palace, the minor and malefic stars, and the natal Four Transformations on top.

Counting only the major-star layer, the 262,980 moments collapse into just 720 distinct charts. That is a 99.726% collision rate. On average one chart is shared by 365.25 moments, and the most crowded chart is shared by 602.

720 is not an arbitrary count, it is a closed-form ceiling. The Life Palace lands on one of twelve branches, the Five Element Class is one of five, the star Zi Wei lands on one of twelve palaces. Multiply the three degrees of freedom and you get 12 times 5 times 12, which is exactly 720. I checked separately that all 720 combinations actually occur, none missing, and none of a 721st combination has appeared in the forty-five years since 1979. At this layer, the resolution ceiling is welded shut.

Add the Body Palace, the minor and malefic stars, and the natal Four Transformations, and the same 262,980 moments now collapse into 225,754 distinct charts. The collision rate drops to 14.155%. On average one chart is shared by only 1.16 moments, and the most crowded one is shared by just 5. Of the 225,754 charts, 194,498 map to exactly one moment, 25,718 are shared by two moments, 5,206 by three, and the remaining 332 are shared by four or five.

Put the two numbers side by side, 720 against 225,754, a gap of 313 times. Neither number is wrong, they answer two different questions at two different resolutions. The major-star layer asks what the skeleton of a chart looks like, and skeletons naturally come in few shapes. The strict layer asks where every minor star, malefic star, and miscellaneous star lands on top of that skeleton, and adding variables blows the count up exponentially.

The Bazi classic Ditian Sui, in Ren Tiechiao's annotated edition, raises the identical puzzle from the other tradition. It attributes the divergence among people born at the same hour to finer gradations of the hour itself, to differences in geography, and to differences in family standing. Translated into Zi Wei terms, the first factor is something current charting tools can already refine with a true solar time correction. The other two, location and family background, were never built into this system as inputs in the first place, so no amount of tightening the star-placement rules will recover them.

Two boundaries need stating plainly. This census counts moments, not people. All 262,980 moments are weighted equally, as if a birth were equally likely at any hour of any day, which is not how real births are distributed, since daytime hours produce more births than the middle of the night. So every percentage above describes the space of possible birth moments, not a random sample of a hundred people on the street. Second, a high or low collision rate says nothing about accuracy. This article measures how many distinct outputs this symbolic system can produce, not whether any particular output matches a real life.

Are twins guaranteed to share a chart. Not necessarily. Twins born at the same hour will land in the same one of the 720 major-star charts with near certainty, a 99.7% chance. At the strict layer, the odds drop to about 14%, because the minor stars, malefic stars, and natal transformations usually diverge in some detail, and the system currently has no input finer than the birth hour to tell the two apart.

Is 720 the resolution ceiling of Zi Wei Dou Shu. At the major-star layer, yes. It is a closed-form product of three degrees of freedom, the twelve Life Palace branches, the five Element Classes, and the twelve possible positions of Zi Wei, not an estimate from sampling. All 720 combinations occur, and none new has appeared in forty-five years. Once the minor stars and transformations are added back in, the ceiling jumps to the strict-layer order of magnitude, but I have not found a clean closed form for that layer yet, only the measured figure of 225,754.

Does this data mean Zi Wei Dou Shu is not accurate. No, this article only answers how many distinct charts the system itself can produce, which is a separate question from whether any of those charts fits a real life. A system that produces few distinct outputs is not thereby useless, and one that produces many is not thereby validated.

Can I rerun this myself. Yes, you need a charting engine that casts a chart for each of the twelve double-hours of every day from 1965 to 2024, then tallies distinct combinations under both the major-star and strict definitions. If your numbers do not match, check the transformation school in use, whether true solar time is applied, and how the Body Palace is assigned, since any one of those three will shift the result.

The same line about people born at the same hour has sat in the classics for centuries as an observation without a number attached. Now the number exists. At the major-star layer, 99.7% of people will collide with someone else. At the strict layer, only 14% will. Whether you put any stock in the symbolic system is a separate matter from whether the count itself holds up, and this one is built to be checked by anyone willing to rerun it.