How accurate is Satmarg Astro?

Our planet positions differ from the Swiss Ephemeris by at most 0.43″ for the Sun to Saturn and 5.98″ for the Moon, and tithi and nakshatra end times by at most 0.11 min, measured over samples from 1900 to 2100. For festivals, 441 of the 441 dates Drik Panchang published for 2025 and 2026 are the dates we compute. Every figure below gives its sample size, maximum and median error.

Last verified: 16 October 2026. This page reads the output of an automated test (TestAccuracyReport). It runs again with every engine change and fails while the numbers here are out of date. Engine version 1.3.0.

Sun to Saturn, largest error

0.43″

Moon, largest error

5.98″

Tithi end times, largest error

0.11 min

Festival and vrat dates matching Drik Panchang, 2025–2026

441/441

1″ (one arcsecond) is 1/3600 of a degree; 1′ (arcminute) = 60″. One nakshatra spans 13°20′ = 48,000″.

Planet positions (sidereal, Lahiri)

161 instants spread evenly over 1900–2100, compared with the Swiss Ephemeris 2.10.03, which we run only as a test reference; it is not part of the site.

Lagna, houses and dasha

82 charts at 12 places: Varanasi, Mumbai, Delhi and Chennai, and also Sydney, London, New York, Oslo, Tromsø (inside the Arctic Circle), Quito, Cape Town and Anchorage.

Sunrise, sunset, moonrise

The Sun's upper limb on the horizon, 34′ refraction, sea level: the definition panchangs use. Most of what remains is the refraction model.

Tithi, nakshatra and yoga end times

Each search starts 5 hours late, so the test checks that the engine finds the transition, not that it remembers it.

Sign changes, stations and combustion

Every sign change, station and combustion edge of 2025–2030 is compared: none missing, none extra. Rahu's times look larger because the node moves only ~0.05° a day, so a 10″ position difference becomes several minutes.

Eclipses, 2025–2030

Eclipse type, and whether it is visible from each Indian reference city:

• Eclipse type (total, annular, partial, penumbral) (Worldwide): 28/28
• Eclipse visible or not, 5 Indian cities (Delhi, Mumbai, Kolkata, Chennai, Guwahati): 140/140

Festival and vrat dates vs Drik Panchang

Drik Panchang's published dates were copied by hand from its pages (the read date is in each fixture file), never generated from our engine. 2027 and 2028 use the same rules but are not yet checked against a published list.

Festival muhurats of 2026 in 10 cities (Diwali Lakshmi Puja, Janmashtami and Maha Shivratri nishita, Ganesh Chaturthi madhyahna, Holika Dahan): 50 windows, each city from its own sunrise and sunset. Every day matches Drik; over the 100 start and end times the median difference is 0.00 min and the largest 1.00 min (Drik prints whole minutes).

Muhurat: against Drik Panchang's 2026 New Delhi lists

“Agree” = Drik lists the date and we have a window that day; “Drik only” = Drik lists it and we do not (reasons below); “night” = Drik's only window is at night while the purpose is done by day; “ours only” = we have a window, Drik lists nothing. For annaprashan, mundan and vidyarambh Drik's day pages could be read only for January to early March 2026.

Where we do not give Drik's date, and why

• 13 × Rahu kaal: Drik's window falls inside it; Drik's own note says to avoid Rahu kaal, so we cut it
• 12 × Venus combust (Shukra asta): we time asta by a 10° orb; Drik ends it around 1 February 2026, likely by visibility
• 5 × Bhadra (Vishti karana): we cut it for every purpose; Drik's year lists do not
• 3 × Kharmas (Sun in Sagittarius): North Indian almanacs and families avoid it for griha pravesh; Drik's list does not
• 3 × Vyatipata or Vaidhriti yoga: we apply it; Drik's lists do not
• 1 × The accepted tithi and nakshatra never overlap in our hours
• 1 × What passes every rule is under 10 minutes

What we deliberately do differently

• Rahu and Ketu: the true node, not the mean node. Over 161 instants in 1900–2100 the two differ by up to 1.79°, so a sign change can move by days: in 2026 Rahu enters Capricorn on 25 November by the true node and 5 December by the mean node. We show both.
• Ayanamsa: Lahiri (Chitrapaksha) by its rigorous definition, not a linear rate. Raman and KP are available too.
• Ekadashi: the Smarta (householder) date is the main one; where Dashami touches arunodaya, the Vaishnava day is shown separately — 10 of the 49 New Delhi Ekadashis of 2025–2026.
• Muhurat: Bhadra, Rahu kaal, Vyatipata and Vaidhriti are cut for every purpose; kharmas is avoided for griha pravesh too.
• Venus and Jupiter combustion by distance from the Sun (Venus 10°, 8° retrograde; Jupiter 11°), not by visibility.
• Sankranti moments come out about 6 minutes earlier than Drik's (Makar Sankranti 2026: ours 15:07, Drik 15:13). The official Indian Astronomical Ephemeris also gives 15:07: Drik's ayanamsa sits ~20″ above the official value and omits nutation. It changes a date only when the ingress is minutes from sunset.
• Time zone of the birthplace (IANA), historical war-time offsets included — never the device's.

Methodology

• The engine is our own: planets from VSOP87, the Moon from ELP 2000-82B, nodes and ayanamsa computed in code. The Swiss Ephemeris (AGPL) is a development-time reference only: its results were written once as plain numbers (JSON); none of its code or data is in the site.
• Fixtures are never regenerated to make a test pass.
• Median = the middle value when all errors are sorted; max = the largest. Each row gives its sample size and years.
• Every number comes from the Go test TestAccuracyReport (api/internal/astro), which writes web/lib/accuracy/measured.json; this page reads only that file.
• For festivals and muhurat the reference is Drik Panchang, the most cited rule-based almanac; where we differ the reason is written down — we do not force agreement.

Checked against independent sources

Beyond the Swiss Ephemeris and Drik Panchang, the engine was compared with the Government of India's Indian Astronomical Ephemeris (IAE), ISKCON's calendar and MET Norway (10 October 2026). Each row is held by a Go test that fails when the engine changes.

Limits

• The Moon's remaining ≤ 6″ is the ELP planetary-perturbation terms we leave out; a Moon within 6″ of a nakshatra boundary can show the neighbouring nakshatra in another program.
• ΔT after 2026 is a prediction (nobody knows it in advance).
• Sunrise uses a sea-level horizon and fixed 34′ refraction; no elevation correction.
• Muhurat does not check lagna shuddhi, personal tara and chandra bala, Panchaka or regional customs.

Read more

• Our methodology
• The muhurat rules
• Festival calendar
• Why almanacs disagree on festival timings
• Open data (CSV)