Skip to main content
Antikythera mechanism eclipse prediction gearing: corroded bronze wheels and engraved Greek inscriptions in soft raking light against a dark museum backdrop

The Antikythera Mechanism: Ancient Greece's Eclipse Computer

The Antikythera mechanism's eclipse dial ran on the 223-month Saros cycle. See how it worked, what it could not do, and how Saros ties to 2026 and 2027.

Portrait of Maya Okonkwo
Written by

I've seen dark skies firsthand in the Atacama altiplano, Namibia's NamibRand Nature Reserve, and Morocco's Merzouga dunes; every other destination I write about is built from measured Bortle and SQM data, cloud climatology, and primary sources like park services and the International Dark-Sky Association.

Updated Published
A corroded bronze lump from a shipwreck off the Greek island of Antikythera held 30 surviving gear wheels[1], a complexity not seen again until medieval cathedral clocks. One of its dials predicted eclipses using the 223-lunar-month Saros cycle[2]. The cycle has a built-in catch: each repeat is visible from a different part of the globe[3], so a Saros tells you when an eclipse is due, not where to stand.
Length of one Saros
223 months
One Saros equals 223 synodic months[8]
Year it was recovered
1901
Lifted from a trading ship that sank near Antikythera[4]
Known fragments
82
Displayed in the National Archaeological Museum, Athens[26]

A bronze lump that turned out to be a machine

The mechanism was recovered in 1901 from a trading ship that sank near Antikythera[4], and for a while nobody knew what it was. The scrap stayed scrap until 1902, when Spyridon Stais spotted gearing and Greek lettering on it[5]; days later it was being read as an astronomical instrument.
Its designers fused Babylonian and Greek mathematical astronomy in one machine[6], and the eclipse dial is where that fusion is easiest to follow. The dial lists a solar eclipse that took place on 12 May 205 BC[7], an entry that matters more for dating the device than for forecasting.
I find the eclipse dial the most useful part of the machine for anyone who travels for the sky. Moon phases and planet positions are things you can check on any clear night, while an eclipse is the one event where a forecast made years ahead decides whether you book a flight.

How the Saros dial turns months into eclipses

A Saros is 223 synodic months[8], and the mechanism's eclipse dial was built to count in exactly that unit[2]. A synodic month runs from one new Moon to the next, so 223 of them add up to a little over 18 years. The arithmetic behind the dial is Babylonian rather than Greek trigonometry[9], which means the Greeks borrowed their eclipse maths from Mesopotamia and wrapped it in gears.

Studies of the dial show it could report four things about each eclipse[10]:

  • the timing of the eclipse
  • the amount of obscuration
  • the angular diameter of the Moon
  • the position of the Moon at that moment
A second dial, the Exeligmos, tracks three Saros cycles at once, a period of just over 54 years[11]. I would call it the feature that turns a table of eclipse months into a forecast: a month alone is useless to a traveler, while an hour decides which side of the horizon the Sun is on.

0, 8, 16 hours

The extra time each sector of the Exeligmos dial adds to a predicted eclipse time[11]
Put together, the four outputs describe a device that modeled an eclipse rather than merely flagging a date. Obscuration and angular diameter are the same quantities a modern eclipse map uses to separate a total eclipse from an annular one, and the Moon's position is where any path calculation starts. I would not oversell the parallel, but the information the Greek dial carried is recognisably the information eclipse travelers still ask for.

Who built it, and how accurate was it

The maker is unknown. An inscription referring to an athletic competition held on Rhodes makes that island the likely place of origin[12], but I would distrust any guide who names a specific astronomer as the builder.
The date is contested too. The common figure is about 100 BCE[13], though the 205 BC eclipse on the dial has pushed some estimates earlier[7]. A few decades either way changes nothing about what the device did.
Milky Way over the red sandstone mountains of Wadi Rum, Jordan's UNESCO-listed desert stargazing destination

Featured · Destination

Wadi Rum

See the Wadi Rum night sky from a Zalabieh Bedouin camp in Jordan's UNESCO desert: bare tents to glass domes, Bortle 2-3 skies, and year-round stargazing.

The Saros itself is older than the Greek machine. The Chaldeans knew it as a period when lunar eclipses seem to repeat, and it applies to solar eclipses as well[14]. The Antikythera builders put a tested cycle into gears.
Accuracy is best judged against the maths it copied. Freeth, who decoded the glyphs and eclipse inscriptions in 2014[15], fitted a model derived from Babylonian System B to the glyph times and got a total rms error of 1.7 hours, with 1.4 hours on the lunar glyphs and 1.9 on the solar ones[16]. That measures how well the model reproduces the times written on the dial, not how often the dial matched a real sky, and the dial forecast not only the month of an eclipse but also its time of day, magnitude and colour[17]. Ancient-sky claims vary widely in strength, and the 'ancient astronomers' billing at Jabal Ikmah in AlUla is a much weaker one than this decoded instrument.
The 12 May 205 BC entry bears on the date because only one start date fits the dial's solar-anomaly parameter, the first full moon of that May, while every other candidate is more than 10 degrees off[18]. Freeth cautions that the date a dial was designed for need not be the date the machine was built[19]. Evans and Carman had argued for an early date earlier, by other methods[20], and a 2016 proposal from Anastasiou and colleagues to revise the eclipse scheme drew a rebuttal from Freeth in 2019[21]. Treat the dating as live scholarly debate.
The sensible position on the device is a modest one. It is a documented instrument with a decoded eclipse dial, a likely Rhodian origin and a date somewhere between about 205 BC and 100 BCE, and that is plenty to be impressed by without inventing a named genius to credit.

Seeing the mechanism, and planning around the next eclipse

The 82 known fragments are displayed in the National Archaeological Museum in Athens[26]. I haven't seen them in person, so confirm gallery details and opening hours on the museum's own site before you build a day around them.
For eclipse planning the nearest real date is the annular eclipse of 6 February 2027, followed six months later by the August total. Neither needs the Saros; both need a path map, a cloud forecast and a flexible itinerary. If I were choosing between them, I would let the cloud odds along each path decide, because a longer eclipse under overcast skies loses to a shorter one under clear ones.

Recommended tours

Tours are selected for quality, not commission. I earn a small fee if you book — at no extra cost to you.

The people who build whole trips around repeating eclipses are profiled in eclipse chasers and umbraphile culture, and the 2027 night sky calendar lines up the other dates worth traveling for.

Frequently asked questions

What is the Antikythera mechanism?

A geared bronze astronomical device recovered in 1901 from a trading ship that sank near Antikythera[4]. Its fragments held 30 gear wheels, a complexity not seen again until medieval cathedral clocks[1].

Could the Antikythera mechanism predict eclipses?

Yes. Its eclipse dial was based on the 223-lunar-month Saros cycle[2] and forecast an eclipse's month, time of day, magnitude and colour[17].

How does Antikythera mechanism eclipse prediction work?

The dial counts months in Saros units of 223 synodic months[8]. A second dial, the Exeligmos, tracks three Saros cycles over just over 54 years and adds 0, 8 or 16 hours to the predicted time[11].

Who built the Antikythera mechanism?

The maker is unknown. An inscription referring to an athletic competition on Rhodes makes that island the likely place of origin[12].

How accurate was ancient eclipse prediction?

A Babylonian System B model fits the dial's glyph times with a total rms error of 1.7 hours[16]. That is a fit to the written times, not a hit rate against real skies. The Saros has a limit: its repeats are visible from different parts of the globe[3].

Where is the Antikythera mechanism now?

The 82 known fragments are displayed in the National Archaeological Museum in Athens[26]. Confirm gallery details with the museum before you go.

Do total eclipses a year apart share a Saros?

No. The 12 August 2026 eclipse is in Saros 126[23] and the 2 August 2027 eclipse is in Saros 136[24].

Sources

26 references Show
  1. Britannica: Antikythera mechanism (opens in a new tab)

    “Radiographic images revealed that the fragments contained 30 gear wheels, an unprecedented complexity for its time, not seen again until medieval cathedral clocks.”

  2. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “The ancient Greek astronomical calculating machine, known as the Antikythera Mechanism, predicted eclipses, based on the 223-lunar month Saros cycle.”

  3. NASA Eclipse: Saros (opens in a new tab)

    “Because the Saros period is not equal to a whole number of days, its biggest drawback is that subsequent eclipses are visible from different parts of the globe.”

  4. Britannica: Antikythera mechanism (opens in a new tab)

    “It was recovered in 1901 from a trading ship that sunk in the Mediterranean Sea near Antikythera.”

  5. Jones, Ancient Near East Today (opens in a new tab)

    “Stais noticed gears and tiny engraved Greek lettering on the fragments, and within a few days it was realized with excitement that they were the remains of some kind of astronomical device.”

  6. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “its design was a highly ingenious fusion of ancient Babylonian and Greek mathematical astronomy.”

  7. Smithsonian Magazine (opens in a new tab)

    “The eclipse prediction calendar, a dial on the back of the mechanism includes a solar eclipse that happened May 12, 205 B.C.”

  8. NASA Eclipse: Saros (opens in a new tab)

    “One Saros is equal to 223 synodic months.”

  9. Smithsonian Magazine (opens in a new tab)

    “The new date pushes the origin back 50 years or even a century, Markoff writes, and indicates that the math the mechanism uses to predict eclipses is Babylonian arithmetic, not Greek trigonometry.”

  10. Live Science (opens in a new tab)

    “The Antikythera Mechanism could not only predict the timing of eclipses but also reveal characteristics of those eclipses, such as the amount of obscuration, the angular diameter of the moon (which is the angle covered by the diameter of the full moon) and the position of the moon at the time of the eclipse, the study found.”

  11. Nature Humanities and Social Sciences Communications (opens in a new tab)

    “The dial is divided into three sectors, which show eclipse time additions of 0, 8 and 16 h and has a period of just over 54 years (three Saros cycles).”

  12. Smithsonian Magazine (opens in a new tab)

    “But the device also includes an inscription that refers to an athletic competition held in Rhodes, the likely place of origin, experts told the Times.”

  13. Britannica: Antikythera mechanism (opens in a new tab)

    “Dating back to 100 BCE, this device, sometimes called the first known computer, featured the earliest known scientific dials and scales.”

  14. NASA Eclipse: Saros (opens in a new tab)

    “It was known to the Chaldeans as a period when lunar eclipses seem to repeat themselves, but the cycle is applicable to solar eclipses as well.”

  15. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “Eclipses are indicated on a four-turn spiral Saros Dial by glyphs, which describe type and time of eclipse and include alphabetical index letters, referring to solar eclipse inscriptions.”

  16. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “with a lunar rms error of 1.4 hours and a solar rms error of 1.9 hours, giving a total rms error of 1.7 hours.”

  17. Nature Humanities and Social Sciences Communications (opens in a new tab)

    “forecasting not only the occurrence of an eclipse in a particular month but also the time-of-day of the eclipse in hours and characteristics of the eclipse, such as magnitude and colour.”

  18. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “The only matching date is FM1 at-204 May-12 (Table S5), when the mean Sun was at 46.75°: all the other dates are more than 10° wrong”

  19. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “it should not necessarily be inferred that the date of the Antikythera Mechanism is the same as the date for which the Saros Dial was designed.”

  20. PLOS ONE: Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism (opens in a new tab)

    “This surprisingly early date for the Saros Dial has been suggested previously, using conflicting methods”

  21. Nature Humanities and Social Sciences Communications (opens in a new tab)

    “A radical revision of this scheme has been proposed in Anastasiou et al.”

  22. NASA Eclipse: Saros (opens in a new tab)

    “The periodicity and recurrence of eclipses is governed by the Saros cycle, a period of approximately 6,585.3 days (18 years 11 days 8 hours).”

  23. NASA Eclipse: Saros series 126 (opens in a new tab)

    “09566 15 2026 Aug 12 17:47:06 75 329 T 0.8977 1.0386 65.2N 25.2W 26 294 02m18s”

  24. NASA Eclipse: Saros series 136 (opens in a new tab)

    “09568 02 2027 Aug 02 10:07:50 76 341 T 0.1421 1.0790 25.5N 33.2E 82 258 06m23s”

  25. NASA Eclipse: Saros series 136 (opens in a new tab)

    “09528 01 2009 Jul 22 02:36:25 66 118 T 0.0698 1.0799 24.2N 144.1E 86 258 06m39s”

  26. Jones, Ancient Near East Today (opens in a new tab)

    “The 82 known fragments as displayed currently in the National Archeological Museum, Athens.”