Context and explanation of my views here: https://manifold.markets/IsaacKing/by-the-end-of-september-will-manifo
This is a serious market. I'm legitimetly quite confused at why so many people don't see this as an obvious failure of mathematical common sense, and I would appreciate someone helping me understand.
1,000
3.00Just to emphasize what was already said. The 23:59 is a extremely common deadline convention especially for coursework in university. It is the last minute of the given day and helps to avoid any misunderstanding between students and TAs. Since these deadlines are thought to be important and students often wait for the last minutes to do their work I, as a TA, like to err on the side of caution.
If I put the deadline as midnight the submission side will show the date as the "next" day. Now if the system also offers an overview for students for deadline it might decide to only show the date but not the time or a student just glances over the deadline and thinks they have an additional day to do the work. In any case this is not a problem of exact definitions, but a problem of communication. The very fact that 23:59 is such a common convention makes it more necessary to avoid drama.
A related deadline concept is a specific date AoE (Anywhere on earth) which means I can submit up until the day of deadline is still considered today in any timezones. Midnight the next day would be excluded.
@AlexbGoode another way to look at it: we want ranges of the form [x,y) (y exclusive) but humans often interpret ranges as [x,z] (z inclusive). As a workaround we set z=y-c, where c is the smallest unit displayed in the UI.
@AlexbGoode Sounds like the AoE system correctly uses midnight as the cutoff rather than one minute before it, which I think reinforces my point that midnight is the naturally correct time.
I agree that 23:59 can help avoid confusion caused by bad UIs and people who don't know how time works. I just don't see why we need to cater to them. I'd rather encourage people to be better. And in some cases it can cause additional confusion, since if people remember that the thing closes "at the end of the [interval]", they're naturally going to assume that it's actually the end of the interval rather than one minute before it.
@IsaacKing I guess my argument is that the current UI is the expected behavior for many people and therefore it is the best UI decision. If you think that a 'correct' behavior is more important than the expected behavior (for most users) than this market will probably resolve NO, because your desired behavior is, imo, 'correct'.
Side note: The manifold UI also only shows the date and not the time. So the 'bad UI' part applies to this site to some extend.
@AlexbGoode Manifold does have that issue, but only when it's more than a day out. I believe it changes to display in hours once the date gets close.
Someone could reasonably set a close time of 7:00, such as for a market about whether something would happen at night, or about how they wake up. Someone on another continent could pick evening their time, which is early morning our time. Seems better to just train people to check the exact time if it matters to them.
Context and explanation of my views here: https://manifold.markets/IsaacKing/by-the-end-of-september-will-manifo seems to center specifically on This Year.
If for what ever reason the minimum increment is 1 whole minute of time, I posit that 2023-12-31 23:59 is the last minute of This Year. Just as midnight is the begging of each day, technically, the whole minute following 2023-12-31 23:59 would be 2024-01-01 00:00 the start of the following year and therefore does not belong to this year.
@finnhambly points out that midnight can be considered to be ambiguous, to which I say sure, If someone at 9:05 AM asks you what time you went to sleep last night and you reply 12:20 AM, I feel that the general consensus is that you meant the time 8 hours 45 minutes ago, which is technically the same day and it was in the morning, but we often ascribe our sleep cycle to the end of the day rather than to the start of our day.
To respond more generally to the "end of that interval" it somewhat falls within the same logic, if the end of that interval coincides with the start of the following interval and to which interval does the closing time belong, however with the exception of self resolving markets being sniped by whales, I feel like the resolution is either already broadly known and trending to zero or one hundred so there is virtually no profit/market correction to be made, or the market is closing well before a definitive answer can be had and so traders are still operating with imperfect information, in which case I am not convinced that sixty additional seconds of market corrections would be traffic in much volume, say for example in this market where the close is intended to be 90 minutes before the CPI data is published.... theoretically no one should have insider information except possibly the people at the Bureau of Labor Statistics so if it actually closes 91 minutes before the CPI data is published vs 90 minutes before the CPI data is published I don't see it materially affecting people's predictions.
@IsaacKing it is essentially the same as @AlexbGoode's much more succinct post with the interval notation. https://manifold.markets/IsaacKing/by-the-end-of-2023-will-i-believe-t#YhOaKDhb8FjiCkme2rc6
Although sets can overlap, any overlap creates ambiguity so it is desirable to create a partition of time intervals.
@ShitakiIntaki I don't think the overlap matters in the real world; it's not possible to take an action in 0 time.
@IsaacKing but there are no zero units except in theoretical settings, the ambiguity lies in whether the cut off is at the beginning of the unit or the completion of the unit. Like rounding errors in displayed units.
@ShitakiIntaki I think specifically in the context of Manifold Markets, the smallest unit is 1 minute, during which many actions could be taken.
@ShitakiIntaki @IsaacKing actually I don't even know, does Manifold close the market at the start of the unit or the end of the unit? I assume the smallest unit is 1 minute.
@IsaacKing agreed, by definition a specific time has a zero measure, but digital systems work in discrete units.
Even though Manifold uses the HH:MM:SS time format, many use just HH:MM (or am/pm) format in writing (including in comment descriptions and comments) and it can lead to confusion and contradictory descriptions.
The convention to set a deadline to 23:59 is used for coursework deadlines because "Friday 00:00" is confusing, as it's easy to get wrong when you quickly scan it and read "Friday" and "00:00" as midnight on Friday, even though the correct time is midnight on Thursday. There are further failure modes if people write it out in am/pm format, as people may mistakenly write 12am or 12pm rather than "12 midnight" or "12 noon", confusing readers unnecessarily. If you make it clear that it's over at the end of Thursday, it's much clearer. The beginning of Friday — with 00:01 — would also work, but it's a bit weird to be given a deadline that's approximately "before Friday", except the first 60 seconds is fair game.
A minute before is arbitrary but most people write times out to the minute rather than the second, so it makes sense I think :) if it was 23:59:59, people would round to 00:00 and the confusion would be reintroduced!
Even though Manifold uses the HH:MM:SS time format, many use just HH:MM (or am/pm) format in writing (including in comment descriptions and comments) and it can lead to confusion and contradictory descriptions.
My confusion is unrelated to the exact length of the discrepancy. It could be 1 hour before the end of the interval or 1 millisecond before the end of the interval, it wouldn't change my arguments or conclusion.
read "Friday" and "00:00" as midnight on Friday, even though the correct time is midnight on Thursday
No? That's midnight on Friday. Midnight is the beginning of each new day.
What would it even mean for 0:00 to equal midnight, but 0:00 is on Friday while midnight is on Thursday? How could a single time be on two different days at once?
@IsaacKing Ah! I think your certainty in this statement explains a lot:
Midnight is the beginning of each new day.
Your points are valid — 00:00 on Friday certainly is on Friday — but "midnight on Friday" is ambiguous at best (and I believe usually refers to the second after 23:59:59 of that day, rather than 00:00:00 of Friday morning). Googling "which day is midnight on" highlights the confusion here, which I think explains why 23:59 and 23:59:59 are often preferred. Incidentally, the highlighted snippet I get says:
24:00 Sunday or 00:00 Monday are both midnight meaning Sunday to Monday.
which I expect you won't be happy about haha. But most of the results says things like this (which I agree with):
It's a matter of convention, and the informal convention is that "midnight on the 10th" is more commonly the night between the 10th and the 11th. But the term is awfully ambiguous, and people do use it both ways.
(When I've scheduled things with "midnight" deadlines, I always say "11:59pm on the 10th" or something like that, to avoid the ambiguity. From experience, if you don't people will ask which you mean.)
Apparently legal documents often declare that "midnight in reference to a day, means the point of time at which that day ends" (according to Law Insider does, anyway). Regardless of whether this interpretation or your interpretation is better, the fact that people interpret 'midnight' differently means that defaulting to something like 23:59 removes that ambiguity and possibility of confusion.
24:00 Sunday or 00:00 Monday are both midnight meaning Sunday to Monday.
Why wouldn't I be happy about this? There's no mathematical issue with saying that 89:63 Sunday is a valid time that's equal to 18:03 on Wednesday. It's just less confusing to avoid such overlapping times and always default to the smallest representation.
Apparently legal documents often declare that "midnight in reference to a day, means the point of time at which that day ends
Legal documents also say that bees are fish, and that pi = 3.2. You'll excuse me if I don't defer to the law on matters of mathematical truth.
Regardless of whether this interpretation or your interpretation is better, the fact that people interpret 'midnight' differently means that defaulting to something like 23:59 removes that ambiguity and possibility of confusion.
Teaching schoolchildren that pi = 3.2 would also avoid the possibility of confusion about the nature of irrational numbers, and save a lot of time on schoolwork to boot!
Why wouldn't I be happy about this?
Because they label both 24:00 Sunday and 00:00 Monday as "midnight". So they're not just saying 24:00 is a valid time (like 89:63) but that it also has claim to that name.
You'll excuse me if I don't defer to the law on matters of mathematical truth.
Totally agree, the problem here is that it's more about language usage than mathematics. Most people (imo) assume midnight on Friday means "24:00" on that day, rather than 00:00 of Friday morning and — whether you like that or not — you need to be able to account for it if you are to communicate effectively.
Because they label both 24:00 Sunday and 00:00 Monday as "midnight". So they're not just saying 24:00 is a valid time (like 89:63) but that it also has claim to that name.
I think this may be the source of your confusion. "0:00", "24:00", "midnight", "the beginning of the day", "1 hour after 23:00" are all different names for the same time. Just like how "4", "four", "IV", and "1 + 3" are all different names for the same number. How you choose to represent the concept doesn't change anything about the concept itself. 0:00 = 24:00 = midnight, just like 4 = four = IV = 1 + 3. It makes no sense to say that one label applies to the time represented by a second label, but not to the exact same time when represented by a third label. It's the same time no matter how to choose to write it out.
Most people (imo) assume midnight on Friday means "24:00" on that day, rather than 00:00 of Friday morning and — whether you like that or not — you need to be able to account for it if you are to communicate effectively.
Sure, and if people are concerned about that, they can add a clarification in their market description, or pick a different ending time like 18:00. A lot of people also think that 0 isn't even, but that doesn't mean I'd make a website say that just to avoid confusion. If there's a common misconception among the population, we should try to fix the misconception! Not entrench it further.
"0:00", "24:00", "midnight", "the beginning of the day", "1 hour after 23:00" are all different names for the same time.
Haha yeah I understand this, but is kind of unrelated to my point, since the confusion comes when you add the day/date to the time of day.
midnight = 24:00 = 00:00, as you say
midnight last night = 24:00 yesterday = 00:00 this morning ≠ midnight tonight
midnight tonight = 24:00 = 00:00 of tomorrow ≠ midnight tomorrow night
Replacing 'last night'/'tonight'/'tomorrow night' with 'yesterday'/'today'/'tomorrow' in those terms is then much more confusing. So saying "resolves midnight Jan 1 2024" is underspecified (and most would take to mean 24:00:00 Jan 1 2024/00:00:00 Jan 2 2024).
But honestly I'm convinced now that Manifold should just go to the end of the interval, as you believe. I think the 1 min interval is weirdly arbitrary and the confusion caused down-the-line from the closing time pretty minor — particularly since the 24h display of closing times is pretty clear (it's hard to misinterpret "1 Jan 2024, 00:00:00").
I agree "midnight of [day]" is likely to lead to miscommunication, but I think "0:00 on [day]" is unambiguous. People who aren't used to 24 hour time may be a little surprised by it, but I doubt they'd take it to mean the end of that day. You can also de-ambiguify "midnight on Friday" by saying "midnight Friday night", to make it clear you mean the midnight that occurs at the end of Friday.
Glad I could convince you! Now to convince the Manifold admins. :)
Also I'm sorry if my first paragraph in my last comment came across a little condescending, I shouldn't have assumed that you didn't understand that distinction.
@IsaacKing Yeah agreed that "00:00 on [day]" is unambiguous, I thought I'd pick up this argument on the basis that it would get swapped with other things frequently — but realised that wasn't really true as I was writing out examples (for the reasons you just said)!
And don't worry, all good fun!
@finnhambly I'll just note that if the close date display called it "midnight" instead of "0:00", I would find that confusing myself (because I'd be uncertain how the site is using it), and I'd advocate for changing it.