Does astrology work? What accuracy can and cannot mean
Accuracy does three separate jobs in a horoscope, and only one of them can be checked. There is the question of whether the positions are right, which carries a stated tolerance and can be laid beside any other ephemeris. There is the question of whether what a reading describes comes to pass, which this site does not claim. And there is the sensation of reading a paragraph and finding oneself inside it, which is the meaning almost always intended by the word, and the one that owes nothing to the sky.
Three things called accuracy
The first is positional. The sun, the moon and the ascendant are computed from published astronomy, and each figure carries a tolerance stated beside the formula that produced it at /method. That kind of claim can fail, which is the point of publishing it. If a position here disagrees with another ephemeris, the error belongs to this site.
The second is predictive: whether the events a reading describes actually arrive. Nothing here is offered in that sense. Neither the daily pages nor the longer readings state what a coming week will contain, not out of caution but because nothing computed here could support such a sentence.
The third is recognition, the experience of reading a description and finding it exact. This happens between a text and a reader rather than between a reader and the sky. It has been studied for most of a century, and it has a name.
The Forer effect, and why a well-written reading feels personal
In 1948 the psychologist Bertram Forer handed his students what he described as individual personality profiles. Every student received the same paragraph, assembled out of a newsstand astrology column. Asked to rate how well it fitted, the class averaged 4.26 out of five. The psychologist Paul Meehl later attached the name of the showman P. T. Barnum to the phenomenon, and it goes by both names.
The mechanism is not subtle. A sentence true of most people and mildly flattering to accept, sociable in company and privately more careful than anyone realises, invites the reader to supply an instance from a life the writer knows nothing about. The specificity felt in such a sentence was contributed by the person reading it. Written in the second person, it works harder still.
None of this is peculiar to astrology. Personality inventories, letters of reference and eulogies all live on it. Nor does it establish that an interpretation is worthless. It establishes one thing only: the feeling of being described is not evidence that the description is accurate.
What testing an interpretation would require
To be tested, a statement has to be able to fail. That means naming what would count as failure, within what period, past what threshold, and against what base rate, which is how often the same thing befalls someone who consulted nothing at all. A line advising care in close relationships this week satisfies none of the four. A sentence that cannot be wrong cannot be right either, and that is the price such sentences pay.
Attempts have been made. The best known is the double-blind trial Shawn Carlson published in Nature in 1985, in which astrologers were given a natal chart and three personality profiles and asked which profile belonged to the chart. The matches came out indistinguishable from chance, and the design has been argued over ever since.
That result is neither displayed here as a verdict nor explained away. The readings on this site are not written in a form such a trial could grade. Marking as prophecy something never composed as prophecy tells neither side anything.
Nothing here is written for one person
A structural limit is worth stating plainly. A sun-sign column addresses roughly a twelfth of everyone alive. This site computes three positions instead of one, which yields 12 × 12 × 12, or 1,728 combinations: considerably finer, and still, divided evenly among the living, some millions of people to each.
Degrees narrow it further, and a birth instant is shared with nobody. But resolution and accuracy are different quantities. A more particular description simply fails more particularly when it fails.
The case for reading one anyway
If a reading is not a prediction, what remains is a fixed vocabulary applied to a particular week. Two thousand years of the tradition have sorted temperament and conduct into twelve types, and whatever their truth value they are ready-made language for thinking about oneself, much the service a well-posed question in a notebook performs.
On that use, the sentence a reader refuses is worth more than the sentence they accept. Refusal is rarely silent: told that they withdraw under pressure, most people begin at once to explain what they actually do instead. The correction comes from the reader rather than from the page, and of the two it is the more reliable.
The boundaries hold in every direction. Nothing here belongs in a decision about health, money or law. The prose is written by Claude, an AI, in the astrological tradition, while the astronomy is computed rather than generated, a seam /blog/how-horoscopes-are-computed sets out at length. Whether a reading came true is the wrong question to put to it. What it caused a reader to consider is the right one.