Sailors trust the North Star to hold still while everything else wheels around it. It does, for now. But the Earth's axis is not fixed; it leans 23.4° from straight up and swings that lean around a small circle once every twenty-five thousand seven hundred and seventy-two years, the way a slowing top traces a slow cone before it falls. The spin axis points at whatever star happens to lie under it. Five thousand years ago that was Thuban, in Draco. Stellar alignments have been proposed for shafts in the Great Pyramid, built in the 26th century BCE, but their purpose is uncertain and a crooked shaft cannot serve as a sighting tube. Today the North Star is Polaris. In twelve thousand years it will be Vega, the summer beacon, and Polaris will be a vagrant again.
This is precession, and it has two faces. The first you can watch on the sky: the pole walks a circle through the northern stars, handing the title of North Star from one to the next and back again over about 26,000 years. The second is quieter and stranger: the same wobble drags the equinox backward through the zodiac, so that on most birthdays the constellation behind the Sun differs from the sign the horoscope names. Both faces are below, computed live and checked offline. Drag the years and watch the sky turn.
- — The pole's path is a circle of radius 23.4° (the axial tilt) about a fixed point: the pole of Earth's orbit.
- For a stationary star and a fixed axial tilt ε, the minimum distance to the pole circle is
|(90° − ε) − β|, where β is the star's ecliptic latitude. Proper motion changes β; moving stars need a search over time. - Hipparchus discovered precession around 129 BCE from a roughly 2° shift in Spica over about 160 years. A modern-rate illustration over 150 years gives 2.10°, broadly consistent with his rough estimate.
The pole walks a circle
Here is the northern sky drawn the unusual way , centred not on the celestial pole (which moves) but on the ecliptic pole, the one fixed point precession turns around. In this frame the wandering pole traces a perfect bright circle, and the named stars sit nearly still while the centuries run. The crosshair is where Earth's axis points; the nearest of the seven plotted stars is highlighted. Drag the slider, or press play.
The succession of these seven plotted stars reads off the circle in order. Going forward from today the crosshair passes Errai (γ Cephei, ~4000 CE), grazes Alderamin (α Cephei, ~7500 CE), swings wide of Deneb (~7° in this model near 10,000 CE), and arrives at brilliant Vega around 13,700 CE before the long return to Thuban. Polaris and Thuban both pass close to the pole: this model puts Polaris about 0.63° away in 2026 and 0.46° away near 2102, and Thuban about 0.33° away at its ancient closest approach. Most eras make do with a rough pole star. This seven-star sample omits Fawaris (δ Cygni), a magnitude-2.87 star that approaches about 3.22° near 11,546 CE in the long-term comparison described below.
| Star | Closest approach | Nearest it gets |
|---|
All seven rows are computed in your browser by the same model the offline verifier checks; six have published epoch anchors, while Kochab has none in the verifier. The minimum distances are this page's first-order geometry. An assay-line comparison on 14 October 2026 using SOFA's long-term pole and the same linear stellar motion gives Alderamin 1.92° versus this model's 2.20°, and Vega 5.69° versus 4.00°. The differences go in both directions. These are model comparisons, not exact future astrometry; see the apparatus.
Why some stars miss the pole
Look again at the map: in this model the pole circle has a fixed radius ε = 23.4392911°, the obliquity. Hold a star stationary and its distance from the centre is fixed at 90° − β, where β is its ecliptic latitude. Its minimum distance to that circle is then the radial gap: |(90° − ε) − β|, approximately |66.56° − β|. For the moving stars on this page, β changes. The numerical search over years −12,000 to 18,000 finds Polaris at 0.46° near 2102, Vega at 4.00° near 13,679, and Deneb at 6.65° near 10,208. These are minima within this first-order model's search window. The verifier evaluates β at each searched epoch and checks that the radial gap closely approximates the searched distance; it does not turn today's latitude into a permanent limit.
The other face: your sign slid off its stars
The wobble does something else, too. The equinox — the point where the Sun crosses the equator in March, the anchor of the whole calendar of signs — is not nailed to the stars. It creeps backward along the zodiac at about 1° every 71.6 years, a full lap in the same 25,772 years. When Greek and Babylonian astronomers fixed the twelve signs, a little over two thousand years ago, the March equinox sat at the head of the constellation Aries, and sign and constellation agreed. They have been drifting apart ever since.
Today the equinox has backed out of Aries and crossed most of Pisces, where it still lies. It is approaching the Pisces–Aquarius line, which is why people speak of the dawning of the Age of Aquarius, still some centuries off (≈ 2597 CE using IAU boundaries). The practical upshot: on most birthdays the Sun's constellation differs from the sign you were assigned. Roll the wheel below through the centuries, then type a birthday and see the gap.
Two facts the wheel makes plain and most horoscopes never mention. First, the Sun spends wildly unequal time in the real constellations: about 45 days in Virgo but only 7 in Scorpius, then about 18 days in Ophiuchus, the serpent-bearer that the neat twelve-fold scheme pretends isn't on the ecliptic at all. Second, the whole thing is sliding: come back in 2,150 years and the entire correspondence has marched on by about 30°, the width of one equal tropical sign. The real constellations have unequal widths.
How we know: Hipparchus, ~129 BCE
None of this needed a telescope. Around 129 BCE, Hipparchus of Rhodes compared his measurements of the bright star Spica with observations by Timocharis about 160 years earlier. Timocharis had recorded lunar occultations of Spica; Hipparchus measured Spica's separation from the Moon during two lunar eclipses (the eclipsed Moon is approximately opposite the Sun). Spica had shifted by roughly 2° relative to the equinox. Practically all the other stars he compared had moved the same way. He concluded the equinoxes themselves were sliding. In Ptolemy's account, translated by T. L. Heath, his lower bound is expressed as "their displacement in 300 years should not have been less than 3°." That is at least 1° per century, or a full cycle in at most 36,000 years.
The check. As an approximate illustration, apply the modern precession rate (50.29″/yr) over 150 years: the drift comes to , broadly consistent with Hipparchus's rough 2°. This is not the exact historical baseline. His floor of 1°/century is cleared (the modern rate is 1.40°/century), and his implied 36,000-year maximum brackets the model's 25,772. A naked-eye astronomer, two centuries before the common era, with a roughly 160-year baseline, got the third motion of the Earth essentially right.
What this page does and doesn't claim
The model is the first-order one, the picture Hipparchus's successors would recognise: the celestial pole circling a fixed ecliptic pole at a constant 23.44° tilt, the stars carried by linear changes in right ascension and declination. It is a rough picture of the succession, not a guarantee of subdegree distances over millennia. It leaves out two slow refinements: the axial tilt has varied between 22.1° and 24.5° over the last million years, in a cycle of about 41,000 years, and the plane of Earth's orbit drifts as well (planetary precession). The long-term comparison above changes Vega's minimum by about 1.69°. The table is computed from this page's model; the new SOFA comparison uses a different pole with the same approximate stellar motion. The zodiac wheel uses mean solar motion and a 2026 reference; the verifier's rough Aquarius estimate uses a 2000 reference. Neither is an exact constellation-boundary calculation.
The apparatus
Constants and star astrometry: IAU 1980 J2000 obliquity (23.4392911°); general precession about 50.29″/yr, consistent with the IAU 2006 J2000 rate, represented here by a 25,772-year period; SIMBAD ICRS J2000 positions and proper motions for all seven plotted stars; IAU-boundary constellation transits (EarthSky); the equinox's Pisces/Aquarius entries from EarthSky's account of Meeus's calculations. The assay-line comparison uses SOFA's long-term pole routine (Vondrák et al. 2011/2012), with a yearly future search from 2000 to 18,000. Its inputs, method and results are in research/precession-equinoxes/assay-2026-10-14.py and assay-2026-10-14.json. Sources in full: research/precession-equinoxes/SOURCES.md.
node research/precession-equinoxes/verify.mjs → 27/27 checks passed: the period↔rate identity, six anchored pole-star approaches and their succession, the radial-gap approximation at the searched epoch for all seven stars, a modern-rate illustration of Hipparchus's roughly 2°, the 13 constellations summing to a year with Ophiuchus present, the sign-vs-constellation mismatch, and a rough Age of Aquarius estimate against the cited 2597 CE. The page and verifier share the model functions used in these tests; static prose and captions are checked separately.