Navratri is not one schedule. The Tithitattva, quoted by P. V. Kane, gives seven alternative periods for the Durgapuja - from a nine-day count starting at Ashvina Shukla Pratipada down to a single day kept on Mahanavami alone - and the Devipurana lays out a separate sequence of rites from the sixth tithi to the tenth. This page dates all of them for your city, names the text each one comes from, and marks the days the sources genuinely leave undecided instead of quietly choosing one.
Dates computed for New Delhi. A festival day is fixed by which tithi is running at a particular moment of the local day, so the civil date can move with the city - change it above if you are observing elsewhere.
The Tithitattva gives seven alternative Durgapuja periods, and Kane records that most of them find support in the Kalika and other Puranas. All seven are dated below, alongside the Devipurana's separate schedule of rites from the sixth tithi to the tenth and Vijayadashami, which closes the sequence on the tenth under its own aparahna rule. None of the seven is marked as the date, because choosing between them is a choice between traditions rather than a calculation.
Kane records the Tithitattva's seven alternatives for the Durgapuja period and adds that most of them find support in the Kalika and other Puranas. They range in length from the longest form, running from the ninth of the dark half to the ninth of the bright half, through the familiar nine-day count from Shukla Pratipada, a four-day form from Shashthi, an abridged three-day form from Saptami for those who cannot keep nine, a two-day Ashtami-Navami form, and finally Mahashtami or Mahanavami kept alone.
Every almanac you can find online prints one of these and calls it Navratri. The choice between them is real and it belongs to a family, a region or a sampradaya - it is not an output of any calculation. So this page dates all seven, labels each with the alternative number the Tithitattva itself uses, and offers no default, no highlight and no ordering that implies precedence.
Two further rows sit alongside the seven and are not alternatives to them. One is the Devipurana's schedule of rites, described below. The other is Vijayadashami, or Dussehra, which closes the sequence on Ashvina Shukla Dashami and is dated by its own rule: its prescribed time is the aparahna, not the day as a whole, so it is decided on a different division of the day from every one of the seven and can land where the count of nine would not put it.
Alongside the seven fasting periods sits a schedule of rites rather than a period of observance: the Devipurana, quoted through the Kalaviveka, assigns bodhana to the sixth tithi, the entry of the sheaf of plants to the seventh, fast, puja and homa to the eighth, the great worship to the ninth and the leave-taking to the tenth, each joined to a nakshatra.
Where the prescribed nakshatra and the prescribed tithi fall apart, the tithi governs. That is stated twice in the sources this engine follows: the Lingapurana allows the bilva branch to be brought in on the seventh even without Mula, and Devala and Lalla say that even if the nakshatra does not obtain on the specified tithi, the tithi laid down is to be followed. The page therefore dates each item by tithi and reports whether the nakshatra happens to coincide as a bonus, never as a condition.
The nine-day form begins with Ghatasthapana on Ashvina Shukla Pratipada, and a pratipada can be mixed - shared between two civil days. Kane gives two cases verbatim: a pratipada mixed with the amavasya that does not exist the next day must be accepted for starting Navaratra because there is no alternative; and a pratipada that spreads into the next day for less than three muhurtas means the full pratipada is taken. He then writes that for want of space the remaining alternatives are not dealt with.
Those two cases are implemented. Anything outside them is rendered as contested, with both candidate days shown. That is a deliberate refusal rather than a gap: the full start-day matrix would have to come from digests that were never opened, and inventing the rest of the ladder would be this app manufacturing doctrine.
One procedural detail does have a locus and is worth stating: Ghatasthapana is performed by day, not at night, per the Durgarcanapaddhati.
The same lunar month carries two names. In the amanta reckoning of the south and west the month ends at the new moon; in the purnimanta reckoning of the north it ends at the full moon. Bright fortnights carry the same name in both, so Ashvina Shukla is Ashvina either way. Dark fortnights differ by a whole month.
That is why the longest Tithitattva alternative, which starts in a dark fortnight, is filed under amanta Bhadrapada while a North Indian reader would call the same days Ashvina Krishna. Every date on this page prints both names, so you can read it in whichever reckoning your family uses without having to convert anything in your head.
There is a second, rarer way the two names come apart, and it can happen even on a bright-fortnight row. The month printed beside a date is the month running at that day's SUNRISE. When a tithi begins after sunrise - or, for an expunged tithi, never holds at a sunrise at all - the observance is dated on a civil day whose sunrise still lies in the previous month, and the label lags the tithi by a month for that one row. The nine-day Navaratra hits this in 2027. Any row where it happens says so in place of the dark-fortnight explanation.
A Hindu day runs from sunrise to the next sunrise, and which tithi is running at a given moment of that day depends on where you are standing. Move far enough east or west and the tithi in force at sunrise is a different one, so the civil date shifts. This matters most for readers outside India, who are also the readers most likely to be looking a date up.
The city selector at the top of the page is therefore not a convenience, it is part of the computation. Your choice is shared with the daily panchang, the vrat calendar and the Hindu calendar month grid, so you only set it once. Positions are computed to sub-arcsecond precision with the Lahiri ayanamsha.
It gives no puja muhurat. No Ghatasthapana window, no Sandhi Puja window, no auspicious hour for anything. The division of the day named on each row - sunrise, aparahna, pradosha - is the division the DATE was decided on, and reading it as a time to perform a rite would be a misuse of it. Per-festival ritual windows are a separate research programme with their own loci and it has not been done here.
It also dates the Ashvina Navaratra only, the autumn or Sharada one. The Chaitra Navaratra of the spring has no rule in this engine; only its ninth day, Ram Navami, is dated, and you will find it on the main festival calendar.
For educational purposes. Astrological interpretations are traditional and not a substitute for professional advice.