Early epoch data for the Kālacakra calendar tradition
This article looks at a selection of the older epoch data for the Kālacakra karaṇa calendar system. The original epoch for the Kālacakra Tantra calendar is 23 March 806 CE. There will be more about this epoch at the end of this article.
After the tantra itself, the next readily available text quoting an epoch for the Kālacakra calendar is the Kālacakrāvatāra (dus kyi 'khor lo la 'jug pa), by Abhayākaragupta ('jigs med 'byung gnas sbas pa). Although this adds little of interest from an epoch point of view, it is nevertheless an extremely important work, being written by a major Kālacakra expert and having a Sanskrit original still preserved – in the Asiatic Society in Calcutta, India. This text uses an epoch of 1087 CE., but simply calculates the number of years back to the Tantra epoch in 806 CE.; it then uses the epoch data from the tantra.
Then there is a set of three short Indian texts that were written by Śākya Śrībhadra. These three texts are:
Śrī-kālacakragaṇ(an)opadeśa (dpal dus kyi 'khor lo'i rtsis kyi man ngag). This gives the calculations for the main structure of the calendar, the five components (pañcāṅga, yan lag lnga), which depend upon the weekday, anomaly and solar longitude.
(nyi zla 'dzin pa'i rtsis, Sanskrit title not given). This gives the calculation for the longitude of the Moon's nodes and a very brief discussion of the occurrence of eclipses.
Pañcagrahapṛthaggaṇanopadeśa (gza' lnga so so'i rtsis kyi man ngag). This gives the calculations for the longitudes of the five planets.
Śākya Śrībhadra has clearly adjusted the values for the weekday and solar longitude from those given in the tantra, presumably in the light of observation. He has also adjusted the data for Venus. The following is a translation of his calculations for the true month, weekday, anomaly and mean solar longitude. This and other translations in this article are fairly loose; translating properly would entail using the symbolic names that are used for numbers, such as eye (mig) for 2 and season (dus) for six. As the intention here is to explain the calculations rather than produce a literary work, these have all been replaced by the numbers concerned. All the calculations in the texts that are used in this article follow the style of the methods given in the Kālacakra Tantra and Vimalaprabhā – for an example of a full set of these calculations, see this page. In the following, the Tibetan is given ahead of the translation and explanation, and indented:
stong pa rab skyes nas gzung ste // 'das lo byin pas lo dag pa'o // nyi mas bsgyur bas zla bar 'gyur // dgun zla 'bring sogs 'das zla sbyin // gnas gnyis 'og ma mig gis bsgyur // mda' ros thob la gcig bsres te // steng la byin pas zla dag 'gyur //
There are a couple of curious points about this calculation, which can be represented:
m × 1;2 + 1;0
He adds no epoch value for the intercalation index, and yet does add 1 to the true month count. This is very unusual, so why not simply start a month earlier? A first assumption could be that he is simply using the tantra epoch data calculated forwards. Data to match his epoch has been added to the latest version of the open source software, KCK. Using that software we can see that if he is using the tantra data, then the epoch month immediately follows an intercalary month of Kārtikka. So, taking his initial epoch as Pauṣa, he then works back one month to the intercalary month of Mārgaśīrṣa. He seems simply to be using the tantra 806 epoch, without any correction. If he had corrected the data, the intercalary should be a few months later; based on the mean solar longitude of the Sun at the epoch new Moon (see discussion of this here), the intercalation index at his epoch should be 49 rather than zero. But this does not explain why he adds 1 to the true month calculation, instead of starting a month earlier.
Another possible explanation is that he wants the month of Pauṣa to be the second of two months that have an intercalation index of zero – an intercalary month and the month following it have the same intercalation index, either 0 or 1. If this is the case, then the "real" epoch month is actually an intercalary month of Mārgaśīrṣa. So, taking his initial epoch as an easily describable Pauṣa, he then works back one month to the intercalary Mārgaśīrṣa. This may have been done to avoid possible confusion if he had stated "add the months since early-winter (Mārgaśīrṣa, mgo)", as there were two of these. This represents a small difference from the tantra calculation, and is my preferred explanation – the tantra calculations have Kārtikka as the intercalary and give the month of Pauṣa an intercalation index of 2, which would not result from the calculations given here.
zla dag gnas gsum bar ma la // so yis bsgyur te mig gis bsre // 'og la ros thob bar ma sbyang // lhag la mkha' sbyin bar nas gcig // 'og tu phab la drug cus bsgyur // mkha' byin des phri lhag chu srang // bar ma drug cus bgos thob rnams // steng byin lhag ma chu tshod gnas // de la ris bgos lhag gza' gnas //"Write the true month in three places, multiply the middle by 32 and add 2. Divide the lower by 6 and subtract the quotient from the middle. Add a zero to the remainder (ie. multiply it by 10) and carrying 1 from the middle line to the lower, multiply it by 60 and subtract the number to which a zero has been added; the result is pala. Divide the middle by 60 and add the quotient to the top line, the remaider being nāḍī; divide the top by 7 and the remainder is the weekday."
This calculation is straightforward. If we consider the true month count to be zero (counting from the "real" epoch of Mārgaśīrṣa) we get:
0 -> 0
0 × 32 = 0 + 2 – 0 = 2
0 ÷ 6 = 0 rem. 0
This gives an epoch value of 0;2,0 (this is Saturday 2 December 1206.) Calculating from the tantra epoch to this new Moon would yield: 6;56,0. If we consider his stated epoch of the month of Pauṣa:
1 × 32 = 32 + 2 – 0 = 34 – 1 = 33
1 ÷ 6 = 0 rem. 1; 60 – 10 = 50
This gives an epoch value of: 1;33,50. Calculating from the tantra would give: 1;27,50.
gnas gnyis 'og mar klu mig sbyin // dus gnyis (should be "nyi") bgos lhag cha shas gnas // nor la ri dang zla bsres te // steng ma lag pas bsgyur la gzhug // klu lag bgos lhag ril po'i gnas //"Again two places, to the lower add 28 and divide by 126; the remainder is the fractional part; add the quotient and 17 to the top after multiplying it by 2. Divide by 28 and the remainder is the anomaly."
This is also straightforward, and gives an epoch value (Mārgaśīrṣa) for the anomaly of 17:28. The tantra calculations produce the same result.
zla ba dag pa gdengs ka sbyin // gnas gsum bar ma drag pos bsgyur // 'og ma bu ga yon tan gyis // thob pas bar sbyang lhag ma la // mkha' dus bsgyur te thob pa dang // lhag ma drug gis bsgyur te thob // bar mas 'og phab rim sbyangs pa'i // lhag ma chu srang dbugs su gnas // dbus ma drug cus bgos pa'i lhag // chu tshod yin te thob pa la // gcig bsres steng ma lhag (should be "lag") bsgyur gzhug // de skar bgos lhag skar ma'i gnas //"Add 8 to the true month, write in three places and multiply the middle by 11. Divide the bottom by 39 and subtract the quotient from the middle; multiply the remainder by 60 and divide again, multiplying the remainder by 6. Carry 1 from the middle, and subtract the remainders for pala and breaths. Divide the middle by 60 and the remainder is nāḍī; add 1 to the quotient and then add to the top multiplied by 2; divide by 27 and the remainder is mansions."
This is not quite so straightforward. If we calculate for the Mārgaśīrṣa epoch, adding the value of 8, the calculation yields:
8 × 2 = 16 + (1 + 1) = 18
8 × 11 = 88 – 1 = 87 ÷ 60 = 1 rem. 27
8 ÷ 39 = 0 rem. 8; 8 × 60 = 480 ÷ 39 = 12 rem. 12; 60 – 12 = 48 – 1 = 47
12 × 6 = 72, ÷ 39 = 1 rem 33; 6 – 1 – 1 = 4
33 × 13 = 429 ÷ 39 = 11; 13 – 11 = 2
The epoch value is therefore: 18;27,47,4,2. Calculating from the tantra epoch gives: 18;26,55,2,4.
Why does he add 8 to the month count? This takes the calculations back to the new Moon (Sunday 9 April 1206) at the beginning of the month of Vaiśākha. As his data give the mean solar longitude at that new Moon as 1;0,0,0,0, perhaps this was done to simplify the calculations.
The next epoch data he gives, in the second text mentioned above, is for Rāhu; this is 18, the same value as would be derived from the original tantra calculation. In the third text he gives the calculations for the planets, and starts off by again counting back 8 months before establishing his count of days. The epoch data he gives for the planets are therefore for the new Moon at the beginning of the month of Vaiśākha. These data are:
These are all the same as would be calculated from the tantra, with the exception of the figure of 1071 for Venus. The tantra calculations would give 546. There is clearly a major error in the original tantra data for Venus, and we probably have here an attempt to correct for this. However, it was not successful – the figure of 1071 translates to a mean heliocentric longitude of Venus of 171.6°, but the actual mean longitude was 348.9°.
The first of these texts is the "lnga bsdus sgra gcan gza' lnga dang bcas pa'i rtsis gzhi". The following translations are similar to those above, covering the true month, weekday, anomaly and solar longitude:
bsil zer can la zhi ba sogs // 'das pa bsres pa lo dag yin // nyi mas bsgyur la nag sogs sbyin // gnas gnyis 'og ma mig gis bsgyur // 'dod sbyin mda' ros rnyed steng bsre // gnyis 'phel shol yin lhag ma ni // mig gis rnyed pas 'das pa rtogs //"Take the years, Saumya, etc., elapsed since 1248 as the true year; multiply by 12 and add the months since Caitra; write in two places and multiply the bottom by 2, add 13 and divide by 65, adding the quotient to the top; when this increases by 2 there is an intercalary; divide the remainder by 2 to determine the months passed (since the last intercalary)."
This intercalation index of 13 divided by 2 is refered to by Yamaguchi as 6.5. But this is nothing special. It is simply calculated forward from the original tantra epoch, and no change has been introduced. This is easy to check. This text has an epoch year of 1248. So, counting years from the Kālacakra epoch of 806, and then multiplying by 12 and again by 2, we get:
( 1248 – 806 ) × 12 × 2 = 10608
There is no epoch value given in the tantra for the intercalation index, so, nothing to add; next, divide by 65:
10608 ÷ 65 = 163 rem. 13
Divided by 2, this is the figure that Yamaguchi quotes. The text itself gives the figure as 13, for a division by 65; Yamaguchi quotes 6.5 for a division by 32.5 – these are equivalent. In the present example, the epoch value is added after the month value on the lower line has been multiplied by 2; in the example that follows, the epoch value is added before the multiplication. Yamaguchi also states that the author does not "allude to any initial standard year or standard years, but since these were to be found in the bsDus rgyud and 'Grel chen, they were a matter of common knowledge..." This is not the case, and Chogyal Phakpa is using as epoch the year 1248.
steng gi zla dag gnas gsum gzhag // dbus su mig sbyin so yis bsgyur // rtsa bar gcig bsre dus kyis bgo // lhag ma thig les brgyan pa yi // steng nas gcig phab drug cu la // sbyangs lhag gza' yi chu srang yin // thob la gzugs bsre dbus su sbyangs // drug cus bgos lhag chu tshod do // nor la bzhi bsre steng du sbyin // thub pas bgos pa'i lhag ma gza' //"(For the weekday:) take the upper true month and write in three places; add 2 to the middle and multiply by 32; add 1 on the bottom and divide by 6; add a zero to the remainder (ie. multiply by 10) and carrying one from the middle line, subtract from 60 – the result is the weekday pala; add 1 to the quotient and subtract from the middle; divide by 60 and the remainder is nāḍī; add 4 to the quotient and add to the top; divide by 7 and the remainder is the weekday."
Taking a true month count of zero:
0 + ( 1 + 4 ) = ÷ 7 = 0 rem. 5
0 + 2 × 32 = 64; 63 – ( 0 + 1 ) = 62 ÷ 60 = 1 rem. 2
0 + 1 ÷ 6 = 0 rem. 1; 60 – 10 = 50
Epoch value is 5;2,50 (Thursday 5 March 1248).
zla dag gnas gnyis 'og ma la // zla ba mi bdag rnams bsres nas // dus nyis bgos lhag cha shas yin // steng la me bsre mig gis bsgyur // thob pa byin la klu lag gis // bgos pa'i lhag ma tshogs su 'gyur //"(For the anomaly:) write the true month in two places and add to the bottom 161 (yes, really!); divide by 126 and the remainder is the fractional part; add 3 to the top and multiply by 2 and then add the lower quotient and divide by 28; the remainder is the anomaly."
Taking the true month as zero again:
0 + 3 × 2 = 6 + 1 = 7 ÷ 28 = 0 rem. 7
0 + 161 = 161 ÷ 126 = 1 rem. 35
Epoch value is: 7;35
gnas gsum dbus ma drag pos bsgyur // rtsa bar bdun bsre so dgus dbye // lhag ma mkha' ros bsgyur ba la // so dgus phye la rnyed pa des // steng nas gcig phab drug cu la // sbyangs lhag skar ma'i chu srang yin // thob pa gong mar mda' bsres pas // dbus su sbyang bzhin (zhing?) drug cus bgo // lhag ma chu tshod steng ma ni // mig gis bsgyur la thob pa bsre // 'khor los bgos lhag skar ma'o //"(For the Sun:) write the true month in three places and multiply the middle by 11; on the bottom, add 7 and divide by 39; multiply the remainder by 60, divide by 39 and subtract the quotient from 60, carrying 1 from above; the remainder is pala; take the earlier quotient, add 5, and then subtract from the middle; divide by 60 and the remainder is nāḍī; multiply the top figure by 2 and then add the quotient from the middle; divide the result by 27 and the remainder is mansions."
Taking the true month as zero again:
0 × 2 = 0 –1 = 26
0 × 11 = 0 – 1 = –1 – ( 0 + 5 ) = – 6 => 54 ÷ 60 = 0 rem. 54
0 + 7 = 7 ÷ 39 = 0 rem. 7; 7 × 60 = 420 ÷ 39 = 10 rem. 30; 60 – 10 = 50 – 1 = 49
Not described in the text, but, carrying down for breaths and final (13) fractional part:
30 × 6 = 180, ÷ 39 = 4 rem 24; 6 – 4 – 1 = 1
24 × 13 = 312 ÷ 39 = 8; 13 – 8 = 5
This gives an epoch value of 26;54,49,1,5
All these epoch data are precisely those that would be derived by taking the original epoch data in the tantra and calculating forwards to Chogyal Phakpa's epoch of Thursday 5 March 1248 – nothing has been changed or added. Anybody interested in checking all these data for themselves could use the open source software KCK, that is available here.
The next text by the same author is the "dpal dus kyi 'khor lo'i rtsis dus gsal ba'i sgron me". Yamaguchi quotes this text next (although he mistakenly gives it the name of the very short text that precedes it) and indicates the intercalation index of 7:
nam mkha'i dbyings su rab byung sogs // 'das pa bsres pa lo dag yin // nyi mas bsgyur la nag sogs sbyin // gnas gnyis 'og mar ri bsres la // mig gis bsgyur la mda' ro yis // rnyed pa steng byin zla dag yin // lhag ma mig gis rnyed pa yis // 'das pa'i zla lhag nam byung brtag // sum cu gnyis gsum spel 'byung bas // mo 'ongs pa ni shugs las shes //"The years elapsed since prabhava (1267) are the true year. Multiply by 12 and add the months since Caitra. Write in two places; add 7 to the lower value, multiply by two and then divide by 65. Add the quotient to the upper figure to find the true month; divide the remainder by 2 to yield the months since the last intercalary; after 32 or 33 there will be another."
Here is the epoch value of 7, and again, this is easily calculated from the data in the tantra. The epoch for this text is the year 1267, and he essentially divides by 32.5, so we should do the same (the epoch value is added before multiplication by 2):
( 1267 – 806 ) x 12 = 5532
5532 ÷ 32.5 = 170 rem. 7
Again, this is easily calculated from the tantra epoch, and there is nothing special going on – no changes to the intercalation constants, as Yamaguchi suggests, no "secret teaching".
zla ba dag pa gnas gsum gzhag // dbus su gzugs bsre so yis bsgyur // de la nyi ma rnam par sbyin // 'og tu mig bsre dus kyis bgo // lhag ma phyogs kyis bsgyur ba yis // steng nas gcig phab drug cu la // sbyangs lhag gza' yi chu srang yin // thob pas dbus sbyang drug cus bgo // lhag ma chu tshod rnyed steng gzhug // thub pas bgos pa'i lhag ma gza' //"(For the weekday:) write the true month in three places; add 1 to the middle and multiply by 32, then add 12; add 2 to the bottom figure and divide by 6. Multiply the remainder by 10, carry 60 from the middle line, and subtract from this; the result is the weekday pala; subtract the quotient from the middle and divide by 60; the remainder is the count of nāḍī, and the quotient is added to the top, which is then divided by 7, the remainder being the weekday."
M as zero:
M + 0 = 0
M + 1 = 1 × 32 = 32 + 12 = 44 – 1 = 43 – 0 = 43
M + 2 = 2 ÷ 6 = 0 rem. 2; 60 – 20 = 40
The epoch value is 0;43,40 (Sunday 27 March 1267)
gnas gnyis 'og mar nor zla sbyin // dus nyis bgos lhag cha shas yin // steng ma mig gis bsgyur ba la // nor la me bsres sbyin par bya // klu lag gis phye'i lhag ma tshogs //"(For the anomaly:) write the true month in two places and add 18 to the lower and divide by 126; the remainder is the fractional part; multiply the upper value by 2, and add the (lower) quotient after adding 3; divide by 28 and the remainder is the anomaly."
M as zero:
M × 2 + ( 0 + 3 ) = 3
M + 18 = 18 ÷ 126 = 0 rem. 18
The epoch value is 3;18
slar yang zla dag gnas gsum gzhag // dbus ma drag pos bsgyur bar bya // 'og tu nor sbyin so dgus dbye // lhag ma mkha' ros bsgyur ba la // so dgus phye la thob pa yis // steng nas gcig phab mkha' dus la // sbyangs lhag skar ma'i chu srang yin // rnyed pa snga mar ro bsres pas // dbus la sbyangs nas drug cus bgo // lhag ma chu tshod steng ma yin // mig gis bsgyur la rnyed pa gzhug // 'khor los bgos lhag skar ma ste // dhru ba rnam par dag pa'o //"(For the Sun:) again, write the true month in three places and multiply the middle by 11; add 8 to the bottom line and divide by 39; multiply the result by 60; divide by 39, and take the remainder and subtract it from 60 carried from the middle line, the result being pala; the earlier quotient has 6 added, and the result subtracted from the middle; divide by 60 and the remainder is nāḍī; multiply the upper line by 2 and then add the quotient; divide by 27 and the result is mansions, this is the true longitude."
0 × 2 = 0 – 1 = 26
0 × 11 = 0 –1 – ( 0 + 6 ) = 60 – 7 = 53
0 + 8 = 8 ÷ 39 = 0 rem. 8; 8 × 60 = 480 ÷ 39 = 12 rem. 12; 60 – 12 = 48 – 1 = 47
Again, not described in the text, but carrying down for breaths and final (13) fractional part:
12 × 6 = 72 ÷ 39 = 1 rem 33; 6 – 1 – 1 = 4
33 × 13 = 429 ÷ 39 = 11; 13 – 11 = 2
The epoch value is: 26;53,47,4,2
Again, all these data are exactly as would be derived when calculating forwards from the data in the Kālacakra Tantra.
Yamaguchi's next reference is to the text called "rtsis dus gsal ba'i sgron me'i don yang dag par bsdus pa". As its name suggests, this is simply a synopsis of the previous one and naturally uses the same data. The next text by Chogyal Phakpa that Yamaguchi references is the "rtsis kyi gtsug lag dang mthun par nges pa". Yamaguchi claims that this adds a value of 14 (double 7) as the intercalation constant. However, he has misread the text, and it actually adds 17 (thub pa zla ba), which is then doubled anyway, to give an intercalation index of 34. (The following has been updated, May 2013, to add further details, correct a couple of others and correct an earlier calculation error in the anomaly.)
The calculations that he gives are certainly in exactly the same Kālacakra style as those earlier in this article, and yield the same results as the standard Kālacakra system for his epoch day. The only differences are that he starts the calculation from "the first" month of a wood-mouse year and he has a different intercalation index than would be derived from the Tantra calculations.
His epoch is: Wednesday 30 January 1264.
Intercalation index: 34 (Tantra calculations give 3)
Mean weekday: 4;2,10
Mean Sun: 22;45,47,4,2
Why the intercalation index of 34? He is here demonstrating how to derive the Chinese lunar months using Kālacakra methods, and his index is clearly intended to match the Chinese intercalary months. Using his intercalation index, the next intercalary month is in the following year, and is the month Vaiśākha, starting on 17 May 1265. Using a calendar converter on the internet, in the Chinese calendar the following month is intercalary, the fifth Chinese month, starting on 17 June 1265. Using modern calculations, this date is correct as the Sun entered the sign of Cancer on 14 June and entered Leo on 15 July – the first day of the following month. As the Sun did not change sign during the month starting 17 June, that month is therefore intercalary. Given that we cannot be sure that the Chinese calendar to which Chogyal Phakpa would have had access would give exactly the same results as modern calculations, the difference of one month can be ignored – his intercalation index is clearly intended to reproduce the Chinese monthly cycle.
The reason that the Kālacakra calendar is so wrong with the intercalary months is that an approximation is used to determine which months are intercalary. This approximation was valid for about 40 years after the Kālacakra epoch of 806 CE, but by Chogyal Phakpa's time it was hopelessly inaccurate. Unfortunately, this approximation was later used as one of the main foundations of the Phugpa and Tsurphu calendars.
One remaining problem with his calculation is his alignment and numbering of the months. His first month is Māgha in the Indian system, and Tiger in the Chinese. According to the definitions of the two calendars, Māgha (Kālacakra) is the month during which the Sun enters the zodiac sign of Aquarius and in the Chinese Tiger month the Sun enters Pisces. In the month in question, the Sun actually entered Pisces. Regardless of where the error lies, the months are misaligned. All main Tibetan calendars have misaligned the months, and even the present day Phugpa calendar would name what should be Māgha as Pauṣa, what should be Vaiśākha as Caitra, and so forth. As early as the 13th century, the translator Shongton complained that the Tibetans misaligned the Indian names of the months by one month and the great Kasmiri Pandit, Vimalaśrī also complained that the months were named one month earlier in Kashmir. This problem is not confined to Chogyal Phakpa.
We can see the most likely reason for the misalignment in a text by Drakpa Gyaltsen, who was writing several decades before Chogyal Phakpa. His "dus tshod bzung ba'i rtsis yig" is a text on the Chinese calendar, and in it he describes the lunar mansion (often he mentions more than one) in which the full Moon is to be found in eight of the months (the other four are said to follow similarly). In the Indian tradition, this is the source of the names of the months – the mansion in which the full Moon is found in an average month. He writes that in the first month, Tiger, the full Moon is in Māgha (the name he uses is sta pa); in the second month, Rabbit, the full Moon is in Phālguna (the name he uses is khra), in the third month, Dragon, the full Moon is in Citrā (the name he uses is bya'u, which seems to be a Tibetan transcription of the Chinese jiǎo, 角),and so forth; see the table above. (The Indian names of the months are derived from the lunar mansion names, but they are not all identical. For this reason, the month when the full Moon on average is in the lunar mansion Citrā is called Caitra. This only applies to Sanskrit – the Tibetan names of the months are the same as the lunar mansions.)
The problem with this list is that it is potentially very misleading. Since the development of the Dàmíng (大明) calendar (462 CE) by the famous astronomer Zǔ Chōngzhī (祖冲之), the Chinese calendar has taken precession into consideration; it is based, like the Kālacakra, on a tropical system. The months are defined by 24 points around the ecliptic, called solar terms (èr shí sì jiéqì, 二十四節氣 dus gzer nyer bzhi). These are exactly equivalent to points around the ecliptic: 0°Aries, 15°Aries, 0°Taurus, etc. The month during which the Sun passes the solar term equivalent to 0°Aries (the spring equinox) is defined as the second month of the year, Rabbit, and so on. The months are not defined relative to the lunar mansions. The mansions are understood simply as the visible constellations in the fixed stars, and the Chinese made beautiful star maps representing them. Due to precession, the ecliptic and the solar terms move very slowly relative to the lunar mansions – the lunar mansions in which the full Moon is seen in each of the months will therefore slowly change.
The definitions of the months in the Kālacakra system are very similar. The months are defined by 12 points around the ecliptic equivalent to 0°Aries, 0°Taurus, etc. The month during which the Sun passes the point 0°Aries is defined as Caitra, and so forth. The longitude is not actually given in terms of zodiac signs, but in terms of lunar mansions. In this system, the lunar mansions are a division of the tropical zodiac into 27 equal parts – the first lunar mansion starts at 0°Aries – and as a result of precession these slowly move with the ecliptic against the background of the fixed stars. In the definition that is the basis of the Kālacakra calendar, the astronomer is told to determine the time of the winter solstice and set the longitude of the Sun at that time to 20;15 (20 lunar mansions and 15 nāḍī) – this is exactly equivalent to 0°Capricorn. Both Chinese and Kālacakra calendars are based on a tropical zodiac, but the Chinese lunar mansions are sidereal and the Kālacakra mansions are tropical. This is almost certainly the source of the month-misalignment problem.
Taking the Rabbit month as an example, this is the month during which the Sun enters the sign of Aries. On average, the Sun will enter Aries at the time of full Moon, and the Moon will consequently be in the (tropical) lunar mansion that is opposite the first point of Aries; that lunar mansion is Citrā, and not Phālguna as Drakpa Gyaltsen has it. His months seem to be out by one to judge by the lunar mansions that he mentions. His list is clearly not based on any theory of the calendar and tropical lunar mansions – as is any equivalent list regarding the Kālacakra calendar – but is presumably based on observation of the visible lunar mansion constellations. By his time precession would certainly have moved the first point of Aries more deeply into the visible constellation of Pisces, and it seems highly likely that this would result in the listing given by Drakpa Gyaltsen. The lunar mansion at full Moon is not defined in the Chinese calendar as this slowly changes with time; it is something that can only be observed.
Now, as certainly seems the case, if the early Tibetans such as Drakpa Gyaltsen did not understand the nature of precession, and it was assumed that the month in which this list of mansions says the Moon is full in, say, the lunar mansion Māgha is the same as the Indian month of Māgha, then the months will be misaligned, exactly as we later find in the writings of Chogyal Phakpa, and the later Phugpa and Tsurphu traditions of Tibetan calendar. In fact, the month in which the Chinese list says the Moon is full in the lunar mansion Māgha should be identified with the Indian month of Phālguna. It should be remembered that the Tibetans had a Chinese-style calendar long before the Kālacakra system was introduced, and that it would be quite understandable that they would interpret the new system from the point of view of their understanding and experience of the Chinese calendar. To judge by comments made by many early Tibetan writers, they certainly considered the lunar mansions to be identified with the visible fixed stars; in fact many of them still do. Unfortunately, this is fundamentally at odds with the Kālacakra calendar definitions.
Perhaps the best way to illustrate this is pictorially. Given the set of data for the mean longitudes in the tantra, we can then calculate the actual mean positions for other new Moons either side of, and including, the 806 epoch. As an indicator of error, we can take the root mean square (RMS) of the differences between the Kālacakra epoch data and the mean longitudes calculated using modern astronomy (using the formulae in Astronomical Algorithms, by Jean Meeus). This gives us a single number index of accuracy, and larger values of this RMS figure indicate greater divergence from the tantra figures. The following chart gives the RMS figures for new Moons, three years either side of the 23 March 806 (mean) new Moon.
The lowest point, with an RMS value of 5.81, is the 23 March 806 new Moon, which is clearly the best fit to the Kālacakra data by a very long way; it is, in fact, the only reasonable fit. If we were to take into consideration the possibility that the original data were based on an early sidereal zodiac intended to be corrected to equatorial, the match would in fact be closer. And so what of Yamaguchi's proposed epoch of the beginning of "the 16th day of the 1st month of 803" (11 January 803)? Not only does the RMS come out at a high value of 78.08 (any correlation with the tantra data is effectively random), but the Sun is nowhere near the equinoctial point, having a longitude of 295° – in Capricorn!
In an analysis like this we are bound to see peaks and troughs when planets complete their cycles. For example, the two troughs just dipping below 40 in the above chart are times when the mean longitudes of the Sun, Mercury and Mars again come reasonably close to the values in the tantra. (I should point out that in these calculations I have corrected an obvious error in the tantra data for Venus – but using the original data makes little and no significant difference to the results.)
But, have we looked far enough? The following chart shows the same data, but now 50 years either side of the March 806 new Moon.
And for good measure, 200 years either side: