How this panchang is computed

What this build used

These are recommended defaults, not a ruling. Every one of them is a parameter the engine takes, and the value actually used comes back with the answer — which is why the line at the bottom of every page on this site is generated from the computation rather than typed next to it. A method line that can disagree with the numbers above it is worse than no method line.

Where panchangs actually disagree

Sunrise: upper limb or centre. The Sun is half a degree wide and the atmosphere lifts it about 34 arcminutes at the horizon. Taking the moment the first sliver appears rather than the moment the centre reaches the horizon is a difference of roughly four minutes — and since the vara begins at sunrise and rahu kaal, gulika, yamaganda, choghadiya and hora are all cuts of the sunrise-to-sunset interval, four minutes at sunrise is four minutes on everything.

Drik or vakya. Drik means the sky is computed as it is observed; vakya means it is read from the traditional verse tables, which are internally consistent and slowly drifting from the sky. Neither is a mistake — they are two different claims about what a panchang is for — but they are not the same numbers and no page should let you assume they are.

Amanta or purnimanta. Whether a lunar month ends at the new Moon or the full Moon. This does not change a single time on the home page; it changes the name of the month, and therefore the date of a festival. This build names both, because picking one would silently take a side in the one place this whole page exists to avoid taking sides silently. On any day of Krishna paksha the two disagree: the amavasya of Diwali closes amanta Ashvina and falls in purnimanta Kartika, and Janmashtami is Shravana Krishna Ashtami in Pune and Bhadrapada Krishna Ashtami in Delhi on the same afternoon. The calendar draws both as bars →

Adhika masa. A lunar month is 29.53 days and a solar month about 30.44, so roughly every 32.5 months one lunar month contains no solar ingress at all and is intercalary — repeated, and marked Adhika. It is computed from the same rule that names the months rather than from a table, and 2026 has one: Adhika Jyeshtha, 16 May to 15 June. Its mirror, a kshaya month, is detected and deliberately left unnamed; sources compound the skipped names differently and this page will not invent one.

Ayanamsa. The offset between the tropical and sidereal zodiacs, which decides which nakshatra the Moon is in near a boundary. Lahiri, Raman and KP differ by enough to move it. This build uses Lahiri and says so, on every page.

Ghati, pala and vipala are not on this list. Every choice above is a genuine fork where traditions disagree and a page has to declare a side. The Vedic-time reading on the home page is the opposite case: a ghati is a fixed sixtieth of the sunrise-to-sunrise day, a pala a sixtieth of a ghati, a vipala a sixtieth of that — arithmetic on a ratio, not a convention, so it needed no ruling before it could be shown.

How a festival gets its date

A festival is a tithi plus a moment of the day it has to be running at, and the moment is where published calendars actually diverge. A tithi runs 19 to 27 hours against a 24-hour day, so the one running at sunrise is routinely not the one running at midnight — which is why a calendar that dates everything at sunrise is a day out on the festivals anchored elsewhere. These are the six this build computes, and what each one means here:

Pradosha is the one worth arguing about, and it was settled by measurement. Six ghatikas, three muhurtas and “2 hours 24 minutes” are the same measurement — and they are 2 h 24 m only if the muhurta is the fixed 48-minute one. Read as a variable night muhurta, three of them are the first fifth of the night. Lakshmi Puja 2026-11-08: the fifth gives 17:31–20:08 in New Delhi and 16:55–19:29 in Kolkata, both within a minute of the published times; the fixed reading ends 19:55 and 19:19, wrong at both latitudes. So this build takes the fifth, and the reason is a number and not a preference.

One festival, one day. Because a tithi can hold the same moment on two consecutive days — Dashami covers aparahna on both 20 and 21 October 2026 — the honest matcher reports the festival twice. The observance is on the first of them, which is what every published date says for every such case here. Only an unbroken run of calendar days is collapsed: the monthly vrats legitimately recur inside one month, and there are two Ekadashis and two Pradosh Vrats in most of them.

A skipped tithi is taken on the day it runs. A short tithi can begin after one sunrise and end before the next, touching no sunrise at all — dated at sunrise it happens on no day whatever, which is how Chaitra Shukla Pratipada and the lunar new year vanished from 2026. It is placed on the only calendar day any part of it fell in, and the page says the tithi was skipped rather than showing a day whose sunrise disagrees with it.

Where traditions disagree, both are carried. Smarta and vaishnava Ekadashi genuinely fall on different days when the tithi is short — by rule, not by error — and the calendar states both rather than choosing. The month system is likewise quoted in whatever the sources quote it in and not normalised: a Krishna-paksha rule names a different month in the two systems and picks the same day either way, so Janmashtami is filed as amanta Shravana and Karwa Chauth as purnimanta Kartika, each with the other name beside it.

And one place this build dates against the classical anchor, deliberately. Janmashtami’s puja is kept at nishita, and the classical rule is nishita-vyapini Ashtami — but every published calendar dates the day by Ashtami at sunrise, and dating it at nishita would put this calendar a day off all of them in 2027. The row is dated the way the calendars are and the deviation is written at the row rather than buried. The vaishnava date, which uses Ashtami at sunrise together with Rohini nakshatra, is not dated at all: that needs a nakshatra rule this engine does not yet run, and an undated variant is named rather than guessed.

Every one of these was checked against a published date before it shipped, and a moment that is not implemented refuses the festival rather than falling back to sunrise — the failure this section is really about is not a wrong minute, it is a confident date produced by a rule nobody stated. The calendar draws them →

Why the minutes can be trusted

Saying “today is Krishna Tritiya” needs the accuracy any astrology app has always had. Printing “ends 14:37” needs about a hundred times more, and three ordinary shortcuts each cost more than the number being published:

All four are properties of engine/, which every surface on this network computes from and none of them duplicates. The astronomy is verified against independent authority — JPL Horizons and NOAA — and never against another panchang site. Two sites agreeing proves agreement, not correctness; they may simply share a mistake. Convention is a separate question with a separate check, against the tradition this page declares it followed.

What is deliberately missing

Dur muhurtam needs a per-weekday list of muhurta numbers that differs between sources. Varjyam and amrit kaal need a per-nakshatra fraction table. Nakshatra-anchored festivals — Onam’s Thiruvonam, the vaishnava Janmashtami — need a rule kind this engine specifies and does not yet run, so they are named and left undated rather than approximated by the tithi beside them. Solar festivals and the calendar year need the instant of a sankranti, which is a boundary solve this engine does not expose, plus a ruling per festival on whether an ingress after sunset counts for that day or the next. Two entries used to be on this list and no longer are: the month name, since both systems are computed and named as of the ruling above, and festival dates, which are dated by the conventions declared further up this page.

Each of them is a row a conventional panchang layout has a slot for, and filling a slot by inventing a table and printing the invention to the minute is the one failure this product cannot survive. They are named here instead. When they arrive they will arrive with the tradition they came from written next to them.

Where the computation happens

In your browser. There is no server holding a table of answers and no request carrying your city anywhere — which is why this works with the network off, and why it will keep working for any date you ask it for rather than only the range somebody generated. The city atlas is bundled for the same reason.