All clocks are 30 seconds late

Flooring makes more sense in every case, from years to milliseconds and more. A few reasons:

You want to send a message at exactly 13:00:00, you have a digital clock that doesn't show seconds. What you do is that you watch your clock, and as soon as it goes from 12:59 to 13:00, you press the button. That's also how you set the time precisely by hand. With rounding, that would be 12:59:30, who wants to send a message at precisely 12:59:30 ?

You have a meeting at 13:00:00, you watch the clock to see if you are early or late. With flooring, if you see 13:00, you know you are late. With rounding, you are not sure.

It is common for digital clocks to have big numbers for hours and minutes and small numbers for seconds. If you are not interested in seconds, you can look quickly from afar and you have your usual, floored time. If you want to be precise, you look closer and get your seconds. Rounding the minutes would be wrong, because it wouldn't match the time with seconds. And you don't want clocks with a small display for seconds you may not even see to show a different time than those that don't.

And if you just want to know the approximate time and don't care about being precise to the second, then rounding up or down doesn't really matter.

> This is especially apparent when you're trying to calculate "how much time until my next meeting?", and your next meeting is at noon. If it's 11:55, you would usually mentally subtract and conclude: the meeting is in 5 minutes. That's how I always do it myself anyway! But the most probable estimate given the available information is actually 4'30"!

Ok. But what does it mean for a meeting to start at 12:00 when people don't have clocks that will accurately show 12:00? They'll only know when the time is 11:59:30 or 12:00:30, anything between is just going to be a guess. So it seems to me that the start times would just align to the half-minute offsets instead, and we'd be back exactly where we started but with more complexity.

Interesting take!

Both of your pic examples are wrong though. That digital clock does show seconds and the London clock has its minute hand in between minute mark - showing progres between minute mark if you look closely. This is the same for all analogue clocks.

I apologise for my "but, actually...":

Analogue clocks like the face of big ben are not like digital displays, and whether they "show seconds" in the context of the meaning of this article is not, like digital displays, down to whether there is a dedicated hand.

Unlike digital displays, the largest denomination hand on an analogue clock display contains all of the information that the smaller hands do (depending on the movement in some cases).

The easiest way to realise this is to imagine a clock without the minute hand. Can you tell when it's half-past the hour? You can. The hour hand is half way between the two hours.

Again, it depends on the movement, but it is not out of the question that your minute hand is moving once every second, and not every minute. It is down to the number of beats per unit time for an analogue display as to what the minimum display resolution is (regardless of if the movement is analogue or digital itself).

All universal statements like this are wrong and stem from basic ignorance

> So when it's actually 14:15:45, they'll show 14:15. And when the actual time goes from 14:15:59 to 14:16:00, then that's when your clock changes from 14:15 to 14:16.

No, that's a silly mistake, look at the picture (though much better - video) of the analogue clock to see it's not the case, the minutes hand moves continuously, so isn't at 15 at 15:59

> the meeting is in 5 minutes.

That's not the only question we ask about time. Has the meeting/game already started? You can't answer that with an average value

> for some context appropriate reason) you reply with just hours, you would say it's 11!

No, you reply would depend on the context, which you haven't specified.

> Please someone tell me I'm not crazy

Of course not, just trying to replace one ambiguity with another. Maybe instead come up with a more precise minutes display in digital clocks that adds more info like two dots flashing when it's past 30 sec and only 1 dot when it's before? (In phones you could use color or a few more pixels within all the padding?)

It's not really about flooring or rounding, but whether one thinks of time indices as ranges or moments.

Days, as the author points out, are though of with "flooring", but more accurately it could be said that a date is thought of as a range between the times belonging to the date.

Minutes one can consider as ranges or time indices. There the error comes, in switching the interpretation of a start of a duration to an actual estimate of a point of time index.

I don't think this applies to Elizabeth Tower/Big Ben, as it's an analogue clock and, from footage I can find[0], its minute hand appears to move continuously opposed to in steps. (or at least, not in full-minute steps)

Also, I believe it's wrong to say "the average error would be 0" if rounding to nearest minute. The average offset would be 0, but the average error would be 15, to my understanding.

[0]: https://www.youtube.com/watch?v=eUP3DsiqkzA

I can get behind this idea, however, this sentence is wrong:

"If clocks rounded to the nearest minute instead of truncating, the average error would be 0.”

The negative and the positive error don’t cancel each other out. They are both error. The absolute value needs to be used.

Hä?! A clock shows the period of time we are currently in. A clock only showing hours would for example indicate that we are in the 14th hour of the day, for the entire duration of that hour. That is not an error. Similarly, a hh:mm clock will show the hour and minute we are currently in for the duration of that minute.

No clock can display the exact current moment of time. That would require infinite digits, and even then those will be late, since lightspeed will ensure you recieve the femtoseconds and below really late.

The fallacy is it’s a leap from logic to go from “average error is x” is the same as “is x late”. Seeing the exact transition is often if not more useful than minimizing average displayed error.

This is nitpicking, but the transition from "the average error of a truncating clock is 30 seconds" to "therefore all clocks are 30 seconds late!" is seriously wrong. For one, the median is equal to the mean here, so about half of all clocks are less than 30 seconds late, which is a clear contradiction.

> Basically I'm arguing that rounding for clocks would be more useful than flooring. This is especially apparent when you're trying to calculate "how much time until my next meeting?"

Yet a rounding clock provides no way at all for you to know whether the meeting has already started or not.

Not sure where I've heard this, but an idea that's been stuck in my head is this: We don't look at clocks to see what time it is, we do so to know what time it isn't yet:

Have I missed the bus yet? Can I already go home? Am I late for this meeting? Do I still have time to cancel this cronjob? All questions that a rounded clock cannot precisely answer.

> Hours: That's when mentally, I switch to rounding! At 15:48 I definitely feel like it's pretty much 16:00.

Disagree. I would never round to a full hour, only to nearest 5 minutes.

As far as minutes, for clocks that show discrete minutes, it'd be weird to see the minute-hand snap to the next number and think "oh, it's actually 29 seconds before that number". Seeing the snap motion means you're at :00 seconds.

Besides, for a clock that doesn't show seconds, it really doesn't matter. If you need more precision, you just use a timepiece with the extra precision.

While most analog clocks' minute hands sweep from minute to minute, jumping minute clocks have the issue the article brings up.

Depending on when / where you went to school, you may have had analog jumping minute clocks. The ones we had at one school would "give away" when the minute would change because the minute hand would move slightly counterclockwise before changing to the next minute.[0]

Per reddit, some Swiss Railway clocks had jumping minutes, but I have not seen one in person. [1]

Another school I attended had sweep second and minute hands, but would hold the clock at 59 seconds until it matched the master clock. Depending on the particular clock and how well it was maintained, these could be 5 - 10 seconds off. Seems like nothing as an adult, but as a kid wanting to go home, it seemed like an eternity, especially on the last day of class for the semester.

[0] This video shows how clocks worked at my school: https://www.youtube.com/watch?v=jpU_lG_TPP4

[1] https://www.reddit.com/r/clocks/comments/10714a1/hard_time_f...

Clocks show the time that is _passed/past_, that's the point. Hence flooring is correct approach.

Even when you involve physics, you are seeing the nanosecond past. Not the actual time. (The time that takes light to travel from clock to your eyes, then your brain's processing delay via neurons etc.)

Even the thought of being late is the same way. If something starts at 13:00 and you are not already there when it is 13:00:00.000, you are -by definition- late.

The Clock app in iOS works like this in timer mode. If you start a timer for 10 seconds, you'll see the 10 for half a second, then 9 through 1 for a second each, then 0 for half a second, then it beeps. In practice it's pretty stupid and not really useful, so I wish they'd fix it.

There's a reason why normal countdowns work the way they do. You care about the exact moment it hits zero, and with rounding you lose that.

Clocks are not 30 seconds late.

By this logic, the author must also tell people it's "2024" until July 3, and that we're still in the 20th Century until 2051.

Time intervals aren't floors or rounds but thresholds.

This is most evident above the level of hours, minutes, or seconds: we speak of a new day beginning (though the time at which this occurs has varied culturally and regionally), or a new year, decade, century, or millennium (with their own threshold arguments courtesy the AWOL Year Nil in the Julian/Gregorian calendrical system).

Birthdates are an interesting example. In Western/European cultures, it's typical to count by full years elapsed, such that an infant within its first 12 months is 0 years old. In Asian cultures, the age is given as "1", which is reflected in an occasionally-used Western notion of "being in my nth year". That is, a 49 year old is in their 50th year.

What a clock (or calendar) indicates generally is what time period you are presently in, that is, the second of a minute, the minute of an hour, the hour of a day, day of a month, etc., etc. And what's relevant from this point of view is that the threshold is what matters.

The other problem touched on in this essay is of estimating time remaining to a future threshold, and here, rounding is indeed useful. If you have a 10am meeting and the time presently reads 9:55 am, you have fewer than five minutes to arrive punctually. But that is planning and not timekeeping issue, strictly.

Let me unlock your hidden precision.

The writer makes an implicit assumption you just glance at the clock with no prior info.

Where seconds count, you can watch the clock until the moment it ticks. At that time you have greater precision. And due to your own innate sense of time, that precision decays for a little while (maybe even 30s?) rather than instantly disappearing.

I like to synchronize my clocks this way (when seconds are unavailable). Yes, it does mean I invest up to a minute more to set the time, and it's probably not worth it if you're doing so often (eg. area prone to power outages).

> It would be weird if we rounded for years, months and days, that's for sure. I think most people think of those scales as intervals. In other words, July is a period of time, with a start and an end. So are years, centuries, seasons. We are inside of it or outside.

I feel like my sense of time is different from the author's. While it can be useful to round the current hour/minute on some occasions, the information about which exact segment of the day/hour you're in can also be very useful. I can certainly tell that I ask the question of "when exactly is it going to be 12:00?" far more often than "how many seconds have statistically likely elapsed in the current minute?"

The biggest issue for me is that the precise moment of when one minute/hour transitions into the next is important for people. Like, when coordinating an event or meeting, would you prefer it if your clock indicated the precise moment when 12:59:59 becomes 13:00:00 and told you to start the meeting, or would it be better if the clock instead told you that it was "13ish" and you'd have to wait out ~30 seconds by counting in your head?

This also causes a jarring discontinuity - now clocks with a ticking hour hand appear to run 30 seconds late than clocks without, turning on the digital clock setting to show seconds offsets it, and so on. Some people celebrate New Year's or occasions that happen at a specific time 30 seconds early because they no longer have a strong reference point.

Clocks are used as inputs into decisions. So this doesn't matter as long as the decision process accounts for it. If I see that a meeting starts in 5 minutes, I'm not setting a timer get there in 300 seconds, I'm trying to decide if I have time to fill my water bottle on the way or not.

The post is a useful consideration to think if you are trying to do precise tasks though. See Keskival's post below about intervals vs moments to add some more precise ways of thinking.

I'd be very surprised if a smartphone, syncing itself to a reference, is ignoring seconds. Why would it?

To me, if a clock that doesn't display seconds is displaying 09:30, then I assume the time could be anywhere between 09:30:00 and 09:30:59.

Presumably when setting the time on a device that doesn't display seconds, the seconds are being reset to 0 (i.e. setting time to 09:30 is setting it to 09:30:00), so if someone really cares about sub-minute accuracy they can take this into account.

> If it's 11:55, you would usually mentally subtract and conclude: the meeting is in 5 minutes. But the most probable estimate given the available information is actually 4'30"!

Admittedly I'm being a bit pedantic, but this isn't true. The expectation value might be 4'30", but the time is as likely to be 4'59" as 4'30"; assuming it's 4'30" will simply minimize your expected error.

Does OP realize that people don't looks at 11:55 and think it means 11:55:00, but they think, it is just not 11:56 yet.

Computers and smart devices syncing to NTP doesn't have 30s average error rate, it is the author who has 30s average error rate in clock-reading if he reads a digital clock and think that the clock tells that no seconds has passed since the last time it added one minute.

RMSE of current digital clocks rounding down is roughly 34.5 seconds.

RMSE of proposed rounding to closest is 17.5 seconds.

The new method is roughly 17 seconds more accurate on average.

Does this reduced error make a significant difference? Probably not. There are very annoying practical issues with this proposal though.

Rounding down makes it clear what minute has already passed. If your clock says 3:00, you know the current time is definitely some point after 3:00. If something needs to happen at 3:00, and your clock says 3:00, it is time to start.

With this proposed method, 3:00 could mean anywhere from 2:29:30 to 3:00:30. When do you start? Do you start as soon as your clock says 3:00? Then you're starting 30 seconds early. Do you wait until your clock says 3:01? Then you're starting 30 seconds late.

And how do you set a clock that uses this scheme? If I want to set the clock to 3:00, do I wait until 2:59:30 and then set the clock to 3:00?

The headaches that arise from this scheme are not worth the trivial error reduction.

Centuries are also 50 years late on average... And everyone is on average 6 months older than they say.

Also moving the rounding 1 place down doesn't solve anything. Now your second-resolution clock is on average 500ms late unless you move the decimal 1 place down again.

Somehow I do not think that the Big Ben transitions the minute`s hand by using the floor function from Python.

If you had two rounding digital clocks, one with seconds and one without, they would switch to the next minute 29.5 seconds apart. Everyone would hate that.

What a nonsensical article. I could've used the 30 seconds I skimmed it back.

All of this is made redundant by just... knowing this is how clocks work.

If you don't have a seconds hand/counter, you can't predict how many seconds are left in the current minute. If it's 11:45:01, it's 11:45 without seconds. If it's 11:45:59, it's still inside the minute-long interval known to everyone as 11:45. The clock can always be one second away from flipping to the next minute. I just mentally subtract one minute if I'm actually planning on being somewhere down to the fraction of a minute, and planning what I can do beforehand. Mentally subtracting 30 seconds is not any easier.

I don't see this as even having some kind of gimmicky benefit...

I use a certain trick a lot when cooking, or timing anything with precision of a several minutes:

0) Suppose I am steaming vegetables for 8 minutes

1) I will check what time I started them (say 5:30)

2) I will forget the number of seconds that were on the clock when I started

3) I will pull the vegetables off at 5:38:30

All this does is move when the issue he claims exists presents itself.

If we flip the minute display at the 30, then the same problem occurs because if you start timing from the 20, the minute will change in ten seconds.

This article is just kind of stupid masquerading as smart.

The intended interpretation for clocks IMO is "we are now past this time".

Like a mile marker when you are running down a jogging track.

It doesn't matter if the granularity is one-day or even one-year, like the ball drop in Times Square on New Year's Eve.

FTA: All the above clocks have one thing in common: they don't show seconds

One of those clocks is Big Ben in London. http://www.bigben.freeservers.com/clocmech.html:

“Drive to the hands is by an oblique shaft which drives bevel gears positioned centrally on a gantry above the clock, with four shafts running out to each dial. Because there is no remontoire the hands on the dials advance by 2 seconds every two seconds, i.e. at every swing of the pendulum“

⇒ if you look close enough, you can read the time at intervals of two seconds.

I can imagine a 20 and a half year old in the US arguing this logic when fighting an underage drinking charge.

One can speculate that current clocks are already rounding to the nearest minute instead of truncating, i.e. already accounting for this, relative to some hypothetical clock that's 30 seconds slower.

While you could argue that it's flooring, it feels much more like truncating. We have a inner representation of time that's like yyyymmdd-hhmmss-mmmm and we truncate it to the precision we need for a given task. If we need the month we truncate to yyyymm, and if we need hours we truncate to yyyymmdd-hh.

That's quit intuitive for all kind of time precision, and the author's rounding solution would be somewhat counterintuitive, at least for me.

One wrinkle here is the choice of signed error instead of eg. RMSE; absolute error metrics don't care about rounding vs. flooring. Since the conclusion of the analysis hinges on this choice of error metric, I think some justification or discussion of that choice adds interesting context.

Signed error implies that clock-imprecision "cancels"; that being 20 seconds late to your meeting will be undone by being 20 seconds early to lunch afterwards. I'm not sure if this is quite how people reason about 'lateness' though. Obviously in some applications it makes sense (ex. recording event times of some phenomenon of interest), but in scheduling my daily life that's not really how I think about things.

I think RMSE is closer to how people evaluate scheduling of stuff; it doesn't matter if I'm late or early, both will have notable consequences in unique ways. The ambivalence of the RMSE error to rounding vs. flooring is the intuition I would put behind the counterargument fielded by the author:

> But this is just a convention! Truncating, rounding or even ceiling are all valid, as long as we pick one and stick to it.

I think you can also squeeze some juice out of playing with the error metric further. If you want a 'just-in-time' schedule (ie. you want to never be early to anything) then the flooring clock is superior to the rounding clock. If you want the opposite then a ceiling-ing clock would be the best choice.

> Personally, I would never say that it's 10 if the clock shows anything past 10:30

I like to floor my times to the nearest 15-minute interval (10, quarter-past, half-past, quarter-to), only rounding once I've reached ~10:55

Smart enough to make the proposition, but not smart enough to disprove it. Middling intelligence might be the greatest threat to society.

I had a watchface on my Pebble smartwatch that only displayed time in five-minute intervals. This post is making me question whether it rounded up or not.

I have to assume it did because I didn't start being late for everything.

In case you want to see it: https://pebblestyle.com/watchface/BVmexStjq3j/

I enable seconds on my PC and phone clocks, and basically anywhere it's an option. I enjoy seeing if two messages on IRC were just 3 seconds apart as it will often tell you the second person hasn't actually read the first person's message right before theirs (they were probably writing their messages at the same time). Something not obvious if you just see the minutes.

All dumb, manual, HH:MM clocks, that is, unless the setter had patience and used a minute-crossing boundary to synchronize, reducing the error to ~1 second and resets them frequently to not drift more than a maximum of say 10 seconds. Most dumb clocks are likely to have an error of a minute or more because of cumulative drift and few people correct them until they are wildly wrong, say 5 minutes or more.

Or eschew the entire class of problem of daylight savings and drift with a GNSS- or radio time beacon-controlled clock.

That's only the case if you interpret them as HH:MM:00. The issue is fixed if you interpret them as HH:MM:30 instead.

There was a clock on HN a while ago (which I cannot now find) - it has one mode in which the clock only updates itself every random number of seconds, or ticks at different times within each second, oor ticks according to a MIDI tune as well as it can while staying within 30 seconds of the real time.

So, while you can just set your clock forward 30 seconds and be done with it, there are some variations that someone who likes thinking about this are worth playing with.

[My favorite mode of the referenced clock, sets the clock randomly ahead between 1-5 min. Since you don't know how much your clock is off by, you have to assume your clock is on time, and that will give you an average 2.5 minutes to actually show up on time.]

I agree this is an interesting take and fun to think about. But if clocks rounded to the nearest minute instead of truncating, the average error would be still be greater than 0 I think, and not 0, as the author claims. Assuming a sufficiently large number of measurements taken at random times, I think the average error would be 15 seconds.

This post is fascinating but I think it is more mundane than I realized. I thought everyone knew that when a cheap digital kitchen timer shows a time like “0 H 1 M”, that actually means you have exactly 60 seconds or fewer for its alarm to go off.

I might be wrong, but there was a Prince of Persia game (maybe for Game Boy?) that had the same timer behavior. It would show the time remaining to complete the level in minutes and would jump immediately from “1 minute” to 0.

The iOS podcast app is more sophisticated than that. It normally displays how many minutes remain in the playing podcast episode. But what is cool is that at the 30 second boundary, it will decrease one minute. For example, it is at -24:30 or -24:29 that it will go from saying “25 minutes remaining” to “24 minutes remaining.” Then when 30 seconds are left, it is nice enough to show you a countdown of the seconds.

The argument is a bit like saying decimals are wrong because the one in "10" jumps to one after you counted to 9, which is 90% of the range and not 50%.

But that is the point of using multiple digits. Similarily the point of using multiple units of time is that each of the parts is in it self correct. 15:30 always means it is after the 15th hour and after the 30th minute.

If you want to know it more precisely you either wait till the minute changes or you get a clock with seconds. With the proposed system you would not know if the time is before or after the 30th minute. Also: why stop at the minute? You could also round the hour.. But that would introduce a weird jump and the minute would suddenly be negative 30.

I have about a dozen of the absolute cheapest digital clocks around my house, and every single one, I set it to the next integer minute then waited for 00s to hit the button that makes the : blink. So, even if every other person on earth fails to do this, the average is below 30s just due to me.

Interesting,

When I make word clock, I make it change from "SATU LEBIH DUA" ("ONE PLUS TWO", 01:02:00) to "SATU LEBIH TIGA" ("ONE PLUS THREE", 01:03:00) at 01:02:30.

But if I make digital clock, I make it change from 01:02 to 01:03 at 01:03:00.

On a larger analog clock, you can clearly see the minute hand progress from one minute mark to another.

Summer starts at the solstice, when the day is the longest. But by then half the length of the season already passed with the day approaching that point. So I think the seasons are half a season off (seasonal lag notwithstanding).

Any clock which doesn't have a seconds display is probably not accurate to the second and is also very unlikely to have been set to the exact second. Those 2 factors together mean a good chance the time is > 30 seconds out, making the original point moot.

This author is either an idiot or does not understand the concept of precision.

It's very surprising that this article has engendered so much discussion...in support of what he is saying.

10:01:01 and 10:01:59 are both still 10:01, because a minute is a discrete unit of time. Rounding the minute to the next minute is simply wrong. Similarly, rounding to the next hour is also wrong and nobody does this except apparently the author. (Note: 15:48 is almost 1600 but is not 1600.) People might round to the nearest five or ten minutes for convenience (i.e., when reading an analog clock) but that hasn't been a thing since the switch to digital clocks.

I am really annoyed that it is almost universal to show only minutes on clocks- when so much in my life at least depends on knowing time to the second. Even modern digital devices like an iPhone make it really hard to get a seconds readout, when the cost of doing so should be zero.

Some of the things I need precise seconds for frequently in my daily life: buying tickets to concerts, campsites, and kids camps before they sell out, checking into Southwest flights, sailboat race starts, and starting/joining virtual meetings I am hosting.

Am I that weird that this is something I need at least once a day that nobody else seems to need or want?

If you do care about what second it is, you can still be 59s off: eg. if the "rounded" clock shows 12:30, it could be 12:29:30 all the way to 12:30:29.

The error is reduced on average if you only care how far away from exact minute you are, but you simultaneously never know when it is the exact minute.

The question is really about what's more generally useful: the offset of 30s is usually not important enough, which is why most clocks only show hours and minutes. Where it matters, higher precision is used.

Some of the social constructs would be less meaningful (like New Year countdown) as most watches and clocks would show midnight 30s early.

This blog post is a really good argument. The kicker for me was that we all generally switch to rounding when asked for the hour (without minutes) but don't round when asked for the minute (without seconds). That's an inconsistency for which there doesn't seem to be a good reason.

However, this inconsistency only exists orally. When a device displays the time, it never rounds. Changing the convention would probably be very confusing for a lot of people, at least initially.

For now, I'm convinced that when communicating the minute orally, I should round to the nearest minute, just like I do with hours.

This could be a good allegory for sampling intervals. I’m forever having to explain to people that if they are looking at data every 60 seconds they can take a lot longer to notice a problem than common sense would suggest. Your new information shows up on average 30 seconds late and up to 59.99999 seconds late. If you’re trying to compare the last two samples to detect a problem then you’re going to see it up to 2 minutes after it already started, since it could start 1 ms after the first sample. Particularly in systems where multiple actors are using periodic actions.

"It's just a convention"

This is correct. In fact, in one sense our clocks could be argued to be minutes out, on average: originally time was measured by the sun. But by the sun, the earth doesn't rotate in exactly the same time each day during the year. So a clock assuming each day is the same length (like they all do) accumulates an error with respect to sun time of more than 16 minutes,and then loses it again, in a cycle.

But we just collectively decided that it was simpler not to bother with that. We changed the convention so that time is marked with respect to an average day length

This would be an overengineered version of "I set my watch early so I'm not late" except that they argue that all watches should be set early, making it fundamentally pointless

12:00 meetings now start at 11:59:30 and you're still late

Get a seconds precision watch if you care about seconds level accuracy

While interesting this is just… wrong. Imagine a world where:

- I tell people I’m 50 years old but I’m really just 49 1/2

- My phone says it’s Jan 7th but it’s really afternoon on Jan 6th

- My coworker says they’re a sr engineer because they’ve been told that they’ll definitely be promoted soon

Rounding shouldn’t be applied everywhere. Some things in life are supposed to use a floor function; common sense applies and most folks intuitively know that 1:00pm means ‘between 1:00 and 1:01pm’.

The article is claiming that clocks that only show time truncated to a minute are off by an average of 30 seconds, which just isn't true in practice. When I say or hear 11:30, I assume it can be within 11:30:00 and 11:30:59. Humans rarely need accuracy to the second beyond this, so clocks intended to be used by humans truncate to the minute without causing problems. To say they are "late" is a real leap.

I mean, you can reduce this to absurdity too. A clock truncating to seconds is off by an average of 500 milliseconds! The problem is no one cares in day to day usage which is what human readable clocks are made for.

Well I was bothered by a similar problem when tweaking my StumpWM modeline long time ago.

I wanted to show the seconds in the clock on my modeline, but the modeline has a fixed refresh rate, which I set to 1HZ. Then I realized my clock can never be accurate.

Eventually I chose to hide the seconds to stop myself from having those nightmares, but I think they are coming back after reading this article.

> If it's 11:55, you would usually mentally subtract and conclude: the meeting is in 5 minutes. That's how I always do it myself anyway! But the most probable estimate given the available information is actually 4'30"!

But if you want to say it's M minutes, then it's either 4 or 5, and 4:30 rounds to 5, so it all works out once more! :)

A slightly related thing I think about someone's, is if I look at the time and it's 13:05 and then I do a task and when I finish a task it's 13:10, you could say that took 5 minutes, but because we can't see the seconds each time could be up to a minute out. So a total of 2 minutes error! Quite big if you're timing something small like 5 minutes!

Amusingly, one of my kids gets very frustrated with how I answer what time it is on most requests. Most of my answers will be of what the nearest relevant time is. I will never bother determining what the exact minute is from my watch. As I just can't see the value of that. Even from a digital watch, I don't see the need in giving a majorly different answer to the same question within 5 minutes or so.

Time Cube redux?

https://en.wikipedia.org/wiki/Time_Cube

Analog clocks mostly don't have the problem the author is complaining about since most minute hands move once per second and you can easily see (depending on your eyesight and distance to the clock) that the minute is partially consumed.

I agree though that this is a downside to digital clocks which don't show seconds, though whether the best fix is to round instead of averaging is hard to say.

Another perfect example of wasted disk space and bandwidth on the internet 8-/

Not to mention a waste of the one truely nonrenewable resource, our time...

I would argue that eg 13:00 represents an interval, not a moment. So there is no error. But I appreciate the nerdy level of detail.

I think a good part of this comes from how we often mention time: "It's half past two" is interpreted to mean: "it's at least half past two"

If we round seconds, why not microseconds/etc?

Ticking clocks let us know what time it is "at least" and uses the minutes :00 as the "barrier" Shifting that to :30 only causes more confusion IMO

You said it yourself, the data is truncated. We shouldn't be rounding time one way or the other, just truncating it.

A delightful read, though I disagree. Unless we were going to introduce half-minute units (which decreases the scale of the error, but makes it twice as numerous) then shifting our minute units by a half value doesn't seem to actually solve the problem, doesn't it only shift it to a slightly different location?

Unrelated, but I've always wanted a clock that is randomly fast so that I never know exactly how fast it is.

At worst, maybe I can set a maximum "fastness" setting.

Then peeking at the actual time has minimal value over the long run.

I'm sure this already exists. But I haven't come across it yet.

So this is not about all clocks but more about all clocks that do not show seconds. In that case, sure.

I don't think rounding clocks would be any more useful in the scenario about mentally calculating how many minutes until the meeting at noon, because a rounding clock would show noon 30 seconds earlier, so the average difference between 11:55 and noon in a rounded clock vs a floor clock is still 4.5 minutes.

If you use round instead of floor an error persists. So all clocks that don't show seconds are always wrong.

One often unappreciated and amazing thing about the GPS-on-cellphones era is that nowadays everyone has the exact time. Up to the late 20th century it wasn't trivial to have the right time, in fact there was a three digit phone number you could call to get the right time (up to the minute).

Has anyone ever wondered why in the US the week starts on Sunday, but in Europe and Asia it starts on Monday?

Side note. I’m more bothered by this issue on digital thermostats.

I don’t even know if they’re flooring or rounding. And sometimes half a degree makes all the difference. I do wish they’d do temperatures out to one decimal point.

If I actually have to be on time and it says 11:55, I’m not going to wait 5 minutes. I would wait 4 minutes.

If it’s more fluid, then 5 minutes is a fine estimation.

OP’s suggestion of rounding the minutes wouldn’t change anything except make what I’m doing complex.

TFA fatally misunderstands what clocks actually measure. They measure the amount of time that has elapsed. When we say "it's quarter after noon" what we really mean is that 12 hours and 15 minutes have elapsed since 00:00. The reason clocks don't round up is because that's simply wrong.

A minute hasn't elapsed at xx:xx:36, it has only elapsed after :59

tl;dr The author of TFA doesn't actually understand what clocks measure.

> Is it the same for the current minute of the current hour ? I'm not so sure.

Yes it is. Appointments start at the beginning of the minute, not the average. (And while changing clocks WOULD change that, I'm not sure it makes a lot of sense)

Perhaps, to be more precise, "How should clocks without a second-hand round?"

My computer's clock say "twenty five past ten" and I don't even take offense about the missing "-" in twenty five. It's close enough for me.

> It would be weird if we rounded for years, months and days, that's for sure ...

> Is it the same for the current minute of the current hour ?

> I'm not so sure.

I'm sure - it's the same.

We don’t floor clocks, we usually put them up on the wall..

Just sitting here with my mechanical automatic watch like some kind of renaissance fair re-enactor. Probably off by a few seconds, but probably less than 30.

I definitely start saying I'm $age+1 as I approach that birthday, as it seems more accurate. Round, not floor.

Although I have to say I've not really encountered other people doing this.

>Please someone tell me I'm not crazy

No, it's probably just autism.

I'm not trying to be snarky but I really don't see why this matters. It's just a convention. We are not robots, not everything has to be 100% precise.

This is true across other domains! If trains are spaced exactly 10 minutes apart and you arrive at a random time, the expected waiting time is 5 minutes, for example.

Perhaps we could retitle the link, as it stands a bit click-baity.

Perhaps, average error of clocks is 30 seconds higher because minutes use .floor() nor .round()

What a waste of 30 seconds of my life reading that

I also thought it was interesting when basketball essentially added 1 second to each game by including tenths on the game clock

"It would be weird if we rounded for years, months and days, that's for sure. "

We round years too, otherwise I would be in my 30s.

Worse yet, our calendars are all 12 hours off!

This article tries to redefine the correct time by pointing to clocks that are not precise.

I thought about it as a kid, you're definitely not crazy Victor, and I appreciate that you wrote this piece

Fun scissor. Seems to depend on your personal definition of "error", with respect to time.

Wow, I’ve never read a wronger take in my life, and that’s saying something!

Except analog clocks, remember those?

I enjoyed this article for it's tongue-in-cheek approach. Thanks Victor!

Yup. I have my clocks in the house set 30 seconds ahead to counteract exactly this.

Plenty of nitpicking here, for me this piece was a fun and clever thought-ride.

That’s a bit of a cup half-empty way to think about it, innit

I think truncating down allows us to procrastinate a little more :)

This is like asserting that all calendars are 12 hours late.

People are usually interested in the transients.

It's been 2025 for 6 months already...

All the way to the miliseconds I strongly disagree!

A clock is not showing you what time it is, a clock is showing you what time it most recently was. It is marking time, not embodying time.

This is why I dislike clocks without seconds.

I'm sorry, you are crazy.

I show seconds in windows clock

What about ages?

Now I need to extrapolate this to explain why I am 10-30minutes late every time.

[dead]