Is it better to leave them powered on 24 x 7, or to minimize the time they are powered on.
 
My specific scenario is a USB dock with drives I use for backups.  The jobs run overnight.  Is it better to power it down until evening, or just leave it running even though it won't be used all day.
 
Run time or power cycles, which is worse for them?

I never let my externals run for a long time.  The cooling isn't there.

I always turn my computer off after I'm done working.

I just leave them connected and let the OS sleep routines handle things.

I very rarely turn a computer off because overnight is automated backup time. Plus one acts as a mail/media/file server, which happens to have a current up-time of 184 days. It sleeps but is hardly ever powered off nor are the external drives attached to it.

Besides power usage, the argument for switching drives off is primarily to save wear on the spindle bearings. The argument for not switching them off is that heat cycling is bad for electronics and like lots of electronic circuits, most drives that fail do so as they are turned on.

With SSDs it's an irelevant issue of course.

tlw
... most drives that fail do so as they are turned on.

I suspect there is at least some collection bias in that statistic if it is a real statistic at all. A substantial proportion of drive replacements results from inability to read data from sectors on the platter, and does not result from the spindle grinding to a stop. Most of the time following a successful boot sequence the computer is only accessing the drive intermittently and for only particular data, and running mainly from memory. If a drive "fails" while running, you are not going to notice that it has failed unless the hit is to a system-stopper area of stored data. Next time you try to boot, however, everything that needs to be loaded to memory must be successfully accessed or you find you have suffered a "drive failure" at boot time, which looks like the drive failed when it was turned on.

A lot of read issues occur to read head misalignment. Heat makes for less mechanical resistance in bearings. The 'cold startup read errors' phenomenon is do to a worn positioning mechanism. A warm drive tends to dither out positional errors.
 
You have seen this if you've ever experienced boot errors on a COLD startup - left the unit on for thirty minutes, and then successfully performed a boot. (Then hurriedly backed everything up...)

I always turn off my setup when I'm finished with it. Updates only take a few minutes, so I usually let the computer go through those paces while I'm getting things ready to record.
 
Even when I used only platter drives I never had  problems as the result of powering down. Some reasons in favor of powering down are- Saves energy, keeps the studio cooler. All  time the computer is powered down saves use on all the components. If I'm gone and an electrical storm comes along or a power surge, my computer will be off at the time. Hackers love computers that are on 24/7.

If you're going to be away from the machine for a long period of time, it makes sense to shut it down.
If you're only going to be away from the machine for a short period of time, it makes sense to leave it running.
 
HDs are pretty reliable these days (amazing for the task they perform).
 

My machine stays on 24x7x365.
 
My question is really about the backup drive that sits in the dock via USB3.  It only gets used between 11:00p and 4:00a.  I started powering it down (after clicking Safe Removal) in the morning and powering it on again in the evening.
 
Then I started wondering ... should I even bother?  Am I making things worse?  Or does it even really matter?
 
 

This day in age, the startup cycle is not as aggressive as it use to be. TBH, just letting Windows power up and power down that drive "as it is needed" is the easiest method - basically will lower power consumption and unnecessary heat over time. So yes, you are thinking too much into it.
 
[aside to clarify some of the above comments]
HDD media is essentially an optical flat (they are that smooth), and coated with a thin lubricant in case the head makes contact with the platter (media). The heads are cut (with an ion beam) with an "air bearing" beneath them so they actually fly 1 microinch or so off the platter (the platter pulls air under the head and gives it flight).
 
In the old days, heads actually came to rest on the media (in a "landing zone"). This zone was pitted with a laser so it was not smooth to prevent the heads from sticking (stiction) to the disk. The fact that a lubricate was used made the risk of stiction greater for two reasons... the head slowly accumulated this lubricant over time, and stiction can literally tear the head off the arm (HGA, or Head Gimbal Assembly). So, in the old days, if you had a flaky drive... back it up immediately, since parking the heads runs the risk of yanking them off.
 
Nowadays, the heads are parked off the media, which is essentially a ramp under the HGA arm to lift them clear. Even in the case of a "loss of power" a capacitor is used to give enough EMF (electromotive force) to park the heads as the capacitor discharges. When powered on, the disk is brought up to speed so that the "air bearing" will have flight speed before the heads are unparked.
 
So, in the past, start/stops were detrimental, but today it is not the case.
 
As far as "misalignment," this is not the case. A head only knows where it is by passing over the media (the head itself only knows "on track" not its own location, per se). Each sector actually has a series of bits prior to the "write zone." One set is left of track, followed by a set right of track, then an "on track" set. Only after it passes over these (and knows it is on track) does the head perform a write or read operation in the data sector. The sector markers are actually written to the platter by a separate machine prior to manufacture of the drive (the heads in the drive itself never write to these), so the machine that does the track bits does, in fact, need to know "where it is" but the heads themselves rely on that data already being present.
[/aside]