How far apart are two ephemerides?

Why this study exists

Nearly every astrology program in use — and the chart pages most readers rely on — computes planetary positions with the Swiss Ephemeris, published by Astrodienst. Shymea does not. It reads NASA JPL's DE442 through the Skyfield library, a decision explained on our methodology page. That raises a fair question from anyone who checks a Shymea chart against the software they already trust: do the two agree, and if they differ, by how much and where?

We had not published an answer. This is it, with the table attached so the result can be checked rather than believed.

Method

• Both engines were asked for exactly the same quantity: apparent geocentric ecliptic longitude of date, tropical.
• Ten bodies: Sun, Moon, Mercury, Venus, Mars, Jupiter, Saturn, Uranus, Neptune, Pluto.
• 87 instants, chosen before any results were seen: every fifth year from 1900 to 2035 at three times of day (an equinox midnight, a solstice noon, and a November evening), plus three fixed moments — the Apollo 11 landing, the 2000 epoch, and the day this study was run.
• 870 comparisons in total. Differences are reported in arcseconds. Nothing was excluded after the fact.
• Software: Skyfield 1.53 against JPL DE442; pyswisseph 2.10.03 against the Swiss Ephemeris.

Result

Across all 870 comparisons the median difference is 0.17 arcseconds. The 95th percentile is 2.8 arcseconds. The single largest disagreement anywhere in the set is 26.7 arcseconds — which is 0.0074 degrees, under half an arc-minute.

The number that matters to a reader is simpler. In 870 comparisons the two engines disagreed about a zodiac sign zero times, and disagreed about a whole degree zero times. For charts from 1960 onward — which covers essentially everyone alive — the median difference falls to 0.10 arcseconds and the 95th percentile to 0.74 arcseconds.

Where the differences concentrate, and why

The disagreements are not spread evenly. They are almost entirely the Moon, and almost entirely before about 1930. That pattern has a specific cause, and it is not a disagreement about where the planets were.

Both engines convert your civil clock time into the uniform timescale the ephemeris is built on, a correction called delta-T. Before roughly 1955 delta-T is not calculated but measured, recovered from historical observations of eclipses and occultations, and different authorities extrapolate the record slightly differently. Skyfield and the Swiss Ephemeris use different delta-T models, so for a 1900 chart they are effectively reading the sky about 40 to 48 seconds apart in time.

That explanation is testable, so we tested it. If the cause is a shared offset in time rather than a difference in the ephemeris, then every body at a given instant should imply the same offset once you divide its angular gap by its own speed. It does. Estimating the offset from the Moon and removing it, the remaining disagreement across all other bodies has a median of 0.10 arcseconds and a maximum of 3.1 arcseconds. The Moon simply shows the effect most because it moves fastest: about 0.55 arcseconds every second of time.

Neither model is the wrong one. This is a boundary of what is knowable about historical timekeeping, not an error in either program, and it is why a birth time recorded to the minute in 1905 carries a little less certainty than the same record made today.

What this does and does not tell you

• It does tell you that a Shymea chart and a Swiss Ephemeris chart will place every planet in the same sign and the same degree, for any birth date tested here.
• It does not tell you that either engine's house cusps agree — that depends on the house system and the birth coordinates, which this study did not vary.
• It does not say anything about whether astrology works. It measures two pieces of astronomy software against each other, nothing more.
• It is not a claim to be more accurate than the Swiss Ephemeris. Both are serious modern ephemerides and this study found no practical daylight between them.

The data

The complete 870-row table is published as a CSV, linked directly below this page. Each row carries the instant, the body, both longitudes to six decimal places and the difference in arcseconds. Anyone with either ephemeris can reproduce it. If you run it and get something different, we would like to know: support@shymea.com.