Why two Tamil calendars disagree about the date

A Tamil month is the sidereal sign the Sun is standing in, so the month turns when the Sun crosses into the next sign. Two almanacs can put that crossing at different hours, because they compute the Sun differently. The thirukanitha reckoning uses the Sun of modern astronomy. The vakya reckoning uses the Sun of the classical Tamil tradition, which runs a few hours behind it. When those few hours fall either side of sunset, the first of the month moves by a day.

Neither almanac has made a mistake. A calendar is a rule for naming days, not a claim about the sky, and both rules are older than the argument.

This site computes both. The Tamil month here is reckoned by vakya unless you change it, because that is what most printed Tamil calendars follow, and the other reckoning is one tap away under Settings. Every Tamil date on the site carries the rule that produced it.

The twelve Tamil months are the twelve signs. Chithirai is the Sun in Mesham, Vaikasi the Sun in Rishabam, and so on to Panguni in Meenam. The instant the Sun crosses a sign boundary is the sankramana, the ingress.

A month begins on the day of the ingress when the ingress happens before local sunset. When it happens after sunset, the month begins the next day. That single line is where a few hours turn into a whole day.

Because the rule reads a local sunset, two towns can take the very same instant and land on opposite sides of their own sunsets. Kerala keeps the same shape of rule but draws the line earlier, at the aparahna limit: three fifths of the daylight after sunrise, which falls around one in the afternoon in Kochi rather than a quarter past six. An ingress in the middle of that afternoon starts the Malayalam month a day later than the Tamil one in the same street, off the same Sun.

Vakya means sentence. The classical Tamil almanac tradition carried its numbers as Sanskrit sentences, with the digits encoded in the katapayadi system and memorised rather than computed afresh. The Tamil vakya almanacs descend from the Vakyakarana of about 1300 CE, which stands in the parahita line of Haridatta and uses Aryabhata’s own parameters rather than the Surya Siddhanta’s.

This site implements those parameters directly: a Mahayuga of 1,577,917,500 civil days divided by 4,320,000 solar revolutions, an equation of centre from a 13°30′ epicycle, and the Sun’s apogee fixed at 78°. Nothing there takes an ayanamsa, because the sidereal zero point is already implicit in the epoch and the year length.

Two of those numbers have drifted from the sky. The vakya year works out at 365.2587 days against a true sidereal year of 365.2564, so it is about 3.7 minutes too long, and sixteen centuries of that accumulate. The classical equation of centre reaches 2.149° against the true 1.915°, about twelve per cent too large. Between them they put the vakya Sun behind the real one through most of the year.

The lag is not uniform. It is widest at the Kanya ingress in September, where it reaches about twelve hours, and it closes to nothing around the Mesha ingress in April, where this model puts the vakya Sun up to about an hour ahead. That is why some months agree and others do not.

A lag of a few hours only matters when it carries the ingress across sunset. Most of the time it does not, and the two reckonings print the same first of the month. Over 2025, 2026 and 2027 the two agree in 25 of the 36 Tamil months and differ by exactly one day in the other eleven, which is about a third of the time. In the published almanacs the vakya month is never earlier than the thirukanitha one.

The case this reckoning was built to answer is 23 September 2026. The thirukanitha Kanya sankranti falls at 07:53 IST on 17 September, well before sunset, so thirukanitha makes 17 September Purattasi 1 and 23 September Purattasi 7. The vakya sankranti falls after sunset on the same 17th, so vakya makes 18 September Purattasi 1 and 23 September Purattasi 6. The second is what a Tamil household’s wall calendar says.

The Tamil month here is vakya by default. It is the one default on this site chosen to match what people hold rather than what the sky says. The Pambu panchangam and the wall calendars in most Tamil homes are vakya, and a reader who opens this app beside one of them has to see the same date or the app is simply wrong to them. The thirukanitha reckoning is one tap away under Settings, and it is labelled.

Every other calendar on the site keeps the true Sun. The Malayalam, Telugu, Kannada and north Indian months are all thirukanitha, because their own almanacs are. The engine on its own also falls back to thirukanitha when nothing has chosen, because an engine with no reader should compute the sky rather than a tradition.

The reckoning in force is named beside the Tamil date on the home card and on the panchangam, so you can always see which Sun produced the day in front of you.

The vakya Sun as printed is not one formula but a family of them. Different almanac houses round to different places and correct by hand, so published month starts scatter by two or three hours around any one computation of them. That scatter only shows up as a different date when an ingress falls within a few hours of sunset, which is exactly the case that matters.

Measured against published vakya almanacs, this site’s reckoning reproduces 34 of the 36 Tamil month starts of 2025 to 2027. Those thirty-six dates, the sources each was read from and the two the engine misses are all in the open test file, so the failures can be checked as easily as the agreements. A wider run over eleven years scored 114 of 132 and compared three sets of classical parameters — the Aryabhatan audayika set used here scored best, ahead of the ardharatrika variant and the Surya Siddhanta — but that table was a one-off and is not published with the code, so it is worth less than the thirty-six you can check. Fitting all four parameters freely reached only 122 of 132, and a constant offset bought nothing, so what is left is scatter in the tradition rather than an error waiting to be tuned out.

Where your own almanac differs from this one, its reckoning is the one your family keeps. The point of printing the rule beside the date is not to settle who is right. It is so that a telephone call about the wedding day can be about which reckoning the family follows, rather than about whose calendar is broken.

Your wall calendar and this one, with the rule behind each date.

Why two calendars differ →