Wikipedia talk:Manual of Style/Dates and numbers

It has been 2956 days since the outbreak
of the latest dispute over date formats.

Unofficial anagram of the Manual of Style

Thanks for a small gesture

In all seriousness, I want to thank WikiOriginal-9 for taking the time to reduce the GDQ (gloom-and-doom quotient) of this guideline [1]. EEng 21:58, 12 January 2017 (UTC)

Please refrain from inventing acronyms. Primergrey (talk) 00:26, 13 January 2017 (UTC)

This is your only warning; if you invent a new acronym again, you may be asked to write a date format essay without further notice. Kendall-K1 (talk) 03:50, 13 January 2017 (UTC)

Ha! It's not an acronym – it's an initialism! Na na nuh na na! I WP:WIN! You may now start WP:WHINING! EEng 04:00, 13 January 2017 (UTC)

400 year frequency problem

A reader writes in at WP:OTRS ticket:2017020210016846. This reader released their email with a free license, specifically, "Creative Commons Attribution-ShareAlike 3.0 Unported and GNU Free Documentation License (unversioned, with no invariant sections, front-cover texts, or back-cover texts)".

---

This has always bothered me, and I'm not sure how to fix the issue outside of going in and editing each of the 366 days of the year pages, so I'm hoping you have a way of correcting this easier than I can.

At the top of each of these pages (such as today's - https://en.wikipedia.org/wiki/February_2 ) , wikipedia states the frequency of that day being on a particular day of the week (for a 400 year period). for example, today states

This date is slightly more likely to fall on a Tuesday, Thursday or Saturday (58 in 400 years each) than on Sunday or Monday (57), and slightly less likely to occur on a Wednesday or Friday (56).

While this may be true for some arbitrary selected span of 400 years, if you took a different span of 400 years, the results would be different. This is similar to stating that February 2 always falls on Thursdays (when observing for one year - 2017)

The day of the week in which a specific date falls on rotates through a 28 year rotation. so every 28 years, any day of the year will land on a Sunday 4 times, Monday 4 times, Tuesday 4 times, etc.

If instead of a 400 year period, you had selected a multiple of 28, and used either 392 or 420, you would have found that each day would have fallen equally on each day of the week.

I realize I say "you" when I know that wikipedia is edited by people all over the world, but I think this error should be corrected from each of the date pages as it is incorrect.

---

Anyone with comments should reply here, and I will direct the person to read here. Blue Rasberry (talk) 17:47, 3 February 2017 (UTC)

Maybe put a "citation needed" template after that? But seriously, I don't think this falls under the "simple arithmetic requires no citation" exclusion. Rather than decide whether it's true or not, shouldn't we require an inline source citation? This seems to come from Template:Day, and an entry in the history there says it's from [2]. @Crissov: know anything about this? Kendall-K1 (talk) 17:58, 3 February 2017 (UTC)

FTR our articles are correct. The OTRS OP's premise is mistaken because s/he does not take account of the century leap year rule in the Gregorian calendar.

If you take them into account, then 400 Gregorian years divides into an even number of weeks. But since 400 doesn't divide into 7, there is no way that the weekdays can be evenly split over dates in that period. As our article correctly notes, they are not. Kahastok talk 18:46, 3 February 2017 (UTC)

I would think the result would be a bit different if one were considering the Julian calendar vs. the Gregorian calendar. Jc3s5h (talk) 20:51, 3 February 2017 (UTC)

Maybe, but the day articles like February 2 are explicitly about that date on the Gregorian calendar. Kendall-K1 (talk) 21:28, 3 February 2017 (UTC)

Agree that the Julian cycle would be different (28-year versus 400-year). Disagree that February_2 article is explicitly about that date on the Gregorian calendar, the intro-paragraph assumes such is the case, but the events/births/deaths are a mixture of Julian and Gregorian. We should be more careful, so readers are not disconcerted by the perceived inaccuracies. If we add some footnotes that should help, see detailed suggestion below. 47.222.203.135 (talk) 11:02, 19 February 2017 (UTC)

• It’s probably unexpected information, but infact accurate like Kahastok explained. Unlike Kendall-K1, I do believe that it’s covered by the mentioned exemption. — Christoph Päper 00:03, 5 February 2017 (UTC)

• I don't think it's a big deal, but I don't think it's covered by WP:CALC. There are just too many moving parts to be able to say it "obvious" nothing's been overlooked in the calculations outlined above. If it's right (and I suspect it is) there should be no problem finding an RS. EEng 01:41, 5 February 2017 (UTC)

It's pretty trivial to confirm - look up a calendar for February 1617 or February 2417 and compare them with a calendar for February 2017. If they are the same (and they will be) then the calculation is confirmed.

But I agree. The fact that the OP got it wrong in such a perfectly understandable way demonstrates that it is not a trivial calculation. And the conclusion is somewhat counterintuitive. WP:CALC does not apply to the calculation in those articles without a further source.

That said, it seems to me that the source linked above is sufficient as a source for the text as-is, as only trivial calculation is needed to add together the relevant numbers given in the first table. Kahastok talk 23:10, 5 February 2017 (UTC)

Having done a manual WP:CALC using a spreadsheet (named somewhat ironically 'Calc'), and the detail-tables of the provided source,[3] it is definitely the case that for Groundhog Day under the Gregorian calendrical system, there *is* a 400-year cycle, and it does properly rollover. So for instance 2/2/2000 is a Wednesday, and if you hand-count all the Groundhog Day celebrations from 2000 thru 2399, which is 400 years, there indeed are 58 Tue/Thu/Sat instances, 57 Sun/Mon instances, and 56 Wed/Fri instances, just as wikipedia accurately says (without source albeit) over at February_2.

So here is the flaw in the text OTRS ticket: "While this may be true for some arbitrary selected span of 400 years, if you took a different span of 400 years, the results would be different." Turns out that is NOT a problem, because the Gregorian system (including the proleptic Gregorian calendar which goes back in time indefinitely albeit counterfactually/ahistorically), you can select an arbitrary span of *any* set of consecutive 400-years-duration, and get the same result. Above I used the span from 2000-thru-2399, which started on a Wednesday for 2/2/2000... but if I shifted up a notch, and used the span from 2001-thru-2400, the counterintuitive 58+57+56 outcome would still hold with no changes, because we dropped 2/2/2000 from the span, but replaced it with 2/2/2400, which is also a Wednesday since it is exactly 400 years away, and the Gregorian calendar *has* a 400-year-cycle. As pointed out by Kahastok, this is always the case: 2017 matches 2417, and also matches 1617, and so on, in terms of what the days-of-the-week look like. However, if you go back to 1217, you HAVE to use the proleptic Gregorian, otherwise things will get seriously out of whack mathematically speaking.

Now, there is a different question, which the wikipedia article on February_2 does NOT speak to. The original OTRS ticket says this: "The day of the week in which a specific date falls on rotates through a 28 year rotation." And that *is* in fact the case, for some calendrical systems, as explained at the Solar cycle (calendar) article. As Jc3s5h mentioned above, the result for the Julian calendar would differ, since it uses a different leapday-scheme from the Gregorian. While it is true that February_2 and the other on-this-day articles in wikipedia have an intro-paragraph which only covers the Gregorian calendar, that does not mean wikipedia should ignore the Julian calendar. The very first thing in the February_2#Events section is the Breviary of Alaric which is traditionally dated to February 2nd during 506 Anno Domini. There is no cite for that 2/2/0506 datestamp in the February_2 article, but if you click on the Breviary of Alaric article there is a cite to a couple of sources published in the late 1800s. So the question becomes, were those sources using the proleptic Gregorian calendar, for the year 506 A.D.? Probably they were not, is my guess, although the source doesn't specifically say if they were using old-style (Julian) dates, or new-style (Gregorian) dates.

But regardless of that specific 2/2/0506 question, the February_2 article, and the other on-this-day articles are deficient in two ways. First, it is important to specify the calendar-system being used, for ALL events that are pre-1752-or-so, and certainly for any events that are pre-1582/1583. Such things will usually *not* be in the proleptic Gregorian, they will be Julian dating-style. And second, since at least *one* reader was disconcerted by the claim that there are *always* going to be 58 Tue/Thu/Sat instances, 57 Sun/Mon instances, and 56 Wed/Fri instances of February 2nd during *any* timespan of 400-years, I would suggest that we specify in the lede that this statement is only factual for the Gregorian calendar, and then in a footnote explain that the proleptic Gregorian calendar can theoretically be utilized such that this statement is factual for any 400-year-timespan past or future. And link to an article which explains the math. But the very same footnote should also go on to explain that under *other* calendrical systems, such as the Julian system used for discussion of pre-1582 historical events even nowadays in 2017 (see e.g. October_25#Events which lists the Battle of Agincourt ... despite that battle happening on October 25th Julian 1415ad), the 400-year-cycle is not applicable. In the Julian case, there is instead a 28-year-cycle, and some groups still use the non-Gregorian systems, so for maximum correctness wikipedia ought to have a footnote explaining that the February_2 article and the October_25 article and all 366 other on-this-day articles are *mostly* about the Gregorian date, unless otherwise noted (e.g. 1415 Battle of Agincourt is Julian because the WP:SOURCES don't use proleptic-Gregorian). My thanks to the person who emailed OTRS, and to Bluerasberry for jumping through the copyright-hoops to get it posted on-wiki. 47.222.203.135 (talk) 11:02, 19 February 2017 (UTC)

You are quite welcome. If you will excuse me now, I am running away from this discussion forum. I obviously brought that post to the right place and now I am quite sure that this is a discussion for other people to enjoy. I read all of your post - very interesting and enough for me to stop here. Blue Rasberry (talk) 03:19, 21 February 2017 (UTC)

break for convenience

Yes, I think you're right. The lede says "February 2 is the 33rd day of the year in the Gregorian calendar" but then goes on to list events that, while it doesn't say, I assume are listed by their Julian dates, not (proleptic) Gregorian. We need something like "February 2 is the 33rd day of the year in the Gregorian and Julian calendars" and "dates in this list before 1752 are presumed Julian unless otherwise specified." The exact wording may be a bit tricky. Also I doubt anyone is going to go through and verify all the dates so the wording will necessarily be a bit weasily. Kendall-K1 (talk) 13:37, 19 February 2017 (UTC)

Maybe a footnote could be added to the lead, saying that the article follow the conventions at MOS:JG. Jc3s5h (talk) 15:08, 19 February 2017 (UTC)

Well, it is not usually wiki-honourable to link to the guidelines from body-prose in mainspace  :-)

Which I learned the hard way, by doing it and getting reverted! But I do note that MOS:JG says this: "The dating method used should follow that used by reliable secondary sources (or if reliable sources disagree, that used most commonly, with an explanatory footnote)." So probably the February_2 article, and other on-this-day articles, should follow the example set by WP:RS. What is the tactic most commonly used by On This Day type of publications, that are WP:RS? I would assume they use a mixture of Julian and Gregorian, but they might do something different. In particular, I'd really like to know whether or not the WP:RS utilize the strange-sounding "...but the start of the Julian year should be assumed to be 1 January..." thing that MOS:JG guideline recommends. That would impact our specific use-case, since we are usually listing the year in which an event/birth/death happened. So for instance, George Washington was born "February 22, 1732 [O.S. February 11, 1731]" and we only list him at February_22#Births with 1732 as the year-of-birth. Is that normal/proper/etc? Do some WP:RS crossref him On This Day: February 11th (for G.W. see Feb 22nd), and also, do some RS use 1731 rather than 1732 for Washington's birth-year? Agree that we should get the exact wording hammered out, for footnotes and intro-text, before we go WP:BOLDly changing all 366 entries 47.222.203.135 (talk) 16:01, 19 February 2017 (UTC)

Sources I've noticed, such as Dictionary of American Biography and Dictionary of National Biography always treat January 1 as the beginning of the year, and otherwise use the calendar in force at the time and place of the event. What I don't know is what modern English-language historical publications do when describing an event at a time and place where neither calendar was in force, for example, early 16th century China. Jc3s5h (talk) 22:04, 19 February 2017 (UTC)

"There is no year 0"

Should this be clarified here? It seems it's not widely understood, and at least one of our articles put the year 0 in the mouth of a source that said year 1 until a moment ago. Maybe say Some sources refer to a year 0; where it is necessary to use such sources, try to find a way to work around this (say, using "the beginning of the year" or "the turn of the era", or rounding up to "year 1"), or at least say the converse Do not insert a year 0 if a source uses wording like "turn of the era", etc.? Hijiri 88 (聖やや) 12:44, 12 February 2017 (UTC)

If a source actually says "year 0" is that an obvious mistake we can just fix, or should we use template:sic? Obviously if it's part of a quote we have to use it verbatim, but what if we're paraphrasing? Kendall-K1 (talk) 13:18, 12 February 2017 (UTC)

It's obvious that there is an error, but do we really always know how to fix it? In a lot of cases (population estimates for example) it doesn't matter, but I can't imagine that when sources say "0 AD" they always mean 1 AD -- surely sometimes it's 1 BC, no? That would make correcting it difficult, at least in theory. I don't actually have an answer, though. Ideally we should always be paraphrasing, as a source that makes a mistake isn't the source we should be quoting, but I'm more concerned just with telling our editors not to use "year 0" on Wikipedia. I'm not sure how much detail to provide in the guideline, though. Hijiri 88 (聖やや) 14:01, 12 February 2017 (UTC)

In astronomy the year 0 is 1 BC, see Astronomical_year_numbering. Peter Gulutzan (talk) 14:53, 12 February 2017 (UTC)

My point exactly. If a source uses year 0, we don't know which year they are talking about, and so correcting it is difficult. Hijiri 88 (聖やや) 21:57, 12 February 2017 (UTC)

(edit conflict) When one is exclusively using AD and BC notation (or equivalently, CE and BCE) there is no year 0. But if one is using astronomical year numbering or the version of ISO 8601 adopted in 2004, there is a year 0. Also, in astronomical works, it is common to use several ways of designating time in the same text, and the reader is expected to keep them straight. So unless 聖 tells us exactly which article and which source are at issue, it's impossible to say if there is an error or not. Jc3s5h (talk) 14:57, 12 February 2017 (UTC)

I see that the MOSNUM does permit astronomical year numbering:

• • Astronomical year numbering follows the Common Era and does not require conversion, but the first instance of a non-positive year should still be linked: The March equinox passed into Pisces in year −67.

Jc3s5h (talk) 15:10, 12 February 2017 (UTC)

@Jc3s5h: As I said above, I'm talking about this article. I have to assume there's an indefinite number of others. The lack of a year 0 is not widely understood, hence why scholars like Lawrence Schiffman have to constantly specify it in lectures aimed at the general public. Wikipedia is written by the general public, hence the above-linked error. I am also assuming that there are books containing similar errors. Hijiri 88 (聖やや) 21:57, 12 February 2017 (UTC)

In the article article mentioned by Hijiri 88, there seems to be at least one minor error. However, negative year numbers go with astronomical year numbering. Had the article under discussion been totally correct, it would have been necessary to change "0" to "1 BC", "-10000" to "10001 BC", etc. One would also have to make sure there were no formulas in the article that depended on the years being stated with astronomical year numbering. Jc3s5h (talk) 00:16, 13 February 2017 (UTC)

The article explicitly uses astronomical numbering: "The following table uses astronomical year numbering for dates, negative numbers corresponding roughly to the corresponding year BC (i.e. -10000 = 10,001 BC, etc.)." Changing 0 to 1 has changed an orderly progression into a peculiarly uneven one that implicitly contradicts the preamble. (The actual estimates quoted are for a variety of years circa 0, but they are very broad estimates for values that were changing very slowly; the article doesn't pretend they're precise and we don't need to worry that they should be adjusted for any change between 1BC and 1AD, 0 and 1.) 92.19.26.31 (talk) 22:46, 23 February 2017 (UTC)