Pocket watch movements are comprised of tiny parts like gears, screws, springs, levers, jewels and many other components.
To engineer all these miniature pieces into a running watch was a remarkable accomplishment to begin with, but to be a useful instrument the watch must also keep reasonably accurate time.
These century-old pieces are capable of surprising accuracy, but this is largely dependent on the condition of the balance weights, which are mounted on the rim of the balance wheel and are responsible for positional accuracy. Ideally they still all weigh precisely the same amount and when the watch left the factory the balance wheel would have been perfectly poised, so that no part of the wheel rim was heavier than another. After a century of being worked over by people of varying skill levels, these critical weights have usually been altered, either by swapping them or by grinding part of the heads off, even though there is absolutely no reason to ever have to touch them after the watch was timed at the factory. Once this was done, the poise was ruined and only after much work can it be mostly restored, unless a group weight change is performed and the hairspring re-vibrated.
This is especially true of watches that did not come with mean-time screws, which were used to quickly make bulk timing changes. They are usually found near the balance wheel arms and were made with a special interference thread to prevent them from turning without force. Any company (Elgin was one) that failed to install them on their watches invited a century of hacks to molest the other weights to make timing changes, because when a watch inevitably slowed from thickening oils and had no mean-time screws to make adjustments, the lazy choice was to simply grind off the balance weights to gain time. At some later point the watch was cleaned properly, which then ran too fast because of the missing weight, so washers had to be added to increase mass, all of which ruined the positional accuracy.
In the world of machinery, anything assembled of small parts which is capable of achieving 99% accuracy would be considered very impressive. To put things in a different perspective, if a given watch is only 99.90% accurate it would be off by a minute and a half per day. To break it down further, if a watch is 99.988% accurate it will be within 10 seconds per day, 99.994% accuracy translates to 5 seconds per day, and 99.996% accuracy will result in roughly 3 seconds of error per day. The balance wheel in a railroad-era watch runs at a rate of 5 beats-per-second (BPS), or 18,000 beats-per-hour (BPH), which multiplies out to:
Gravity plays a part in accuracy, which goes back to balance wheel poise and whether the balance weights still all weight the same. Since the heaviest part of the wheel rim will naturally swing down because of gravity, it's obvious that the better the wheel is poised, the less gravity will affect the watch. If the watch is laying dial-up or dial-down then the balance wheel is spinning like a top and the effects of gravity are substantially reduced. Watches that were meant for more exacting requirements of railroad service were adjusted for specific positions, and there were six of them:
Getting a watch to run accurately in all positions took hours of work by a skilled watchmaker after days of careful observations and small subsequent adjustments. Watches that were not adjusted at the factory can still be made to keep excellent time, since they have the same combination of balance weights as the adjusted variants.
Temperature has a big impact on accuracy, not only from the many parts expanding and contracting slightly as they changed shape and dimension, but also from the differing viscosity of its oils. The invention of the earlier bi-metallic balance wheel, or "compensating" wheel, which had cuts in the wheel rim to allow the wheel to change its rate during temperature swings, went a long way toward improving accuracy, as did the invention of alloys like Elinvar that did not react to temperature extremes. Watch factories went to great lengths to test their pieces, usually conducted over the span of several days in climate controlled environments. Movements were carefully timed in temperatures as low as 40 degrees and as high as 90 degrees, and then the balance weights were adjusted, swapped, or re-positioned accordingly by a skilled worker.
The South Bend Watch Co was famous for freezing their watches in ice to demonstrate their accuracy. Several period advertisements survive to show this marketing technique.
The word is Greek for same time and is the accuracy of a watch whether the mainspring is fully wound or almost spent, and in mechanical terms is the ability of a watch to run at the same rate despite any changes in its power source. The red line in this graph represents the drive torque of any mainspring from unspooled to completely wound.
The stopworks, or Maltese Cross, originated in Switzerland in the 1700s as a means of limiting mainspring travel in watches, which evened out its drive torque. This simple mechanism "counted" the number of mainspring barrel rotations, and after a specific number of them it would lock and prevent the mainspring from turning. If a watch had stopworks that were intact and aligned properly, it would then allow the watch to run only in the middle two-thirds of its linear force, shown in green on the graph.
The vast majority of American pocket watches from the railroad era were not equipped with stopworks, but there were other aspects that helped maintain isochronism, like counter-balanced pallet forks, longer hairsprings with overcoils (Breguet springs), and higher jewel counts to help reduce friction.
Watches were meant to be carried, so accuracy is averaged over a 24-hour period of twelve hours in your pocket and twelve hours on your desk, for the two most common positions of dial-up and pendant-up against the three standard measurements:
If the work has been done to spec and everything is within tolerance, then there should be virtually no difference in accuracy between dial up and dial down, since the balance wheel is spinning like a top on either one of its pivots in both of those positions.
Additional timing is available at standard bench rates for the other four positions, but keep in mind that a $15 Casio will spank even the best railroad watch over the long run. In other words, these mechanical antiques are simply not capable of quartz time.
Timing is an average, not a snapshot, since instantaneous positional accuracy is not possible with these antiques. They don't have a recovery mechanism, meaning that if you suddenly shake the watch it will then take a few seconds for the balance to regain equilibrium, and while that's happening the watch is running slower than the rest of the world. I typically time a watch for zero when pendant-up (while it's in your pocket), knowing it will lose slightly from the range of human motion during the day, and a little fast while dial-up (laying on your desk) so it can gain back those few seconds at night.
Static timing is easy, like a clock that just sits there in one position on a shelf in your warm living room with the weights constantly pulling straight down. If I know that a customer does not intend to carry a given piece, but display it running instead in a fixed position, then I will simply time the watch for that one position for greater accuracy.
Watches are susceptible to magnetic fields because many of their parts are ferrous, especially the delicate steel hairspring. When it becomes magnetized the coils can become attracted to the balance arms or to each other, slowing the watch considerably.
It's not only the accuracy that suffers when a watch is run far past the point where the oils have dried out. Crucial components like pivots or bushings can easily become worn down to the point where they will need to be replaced, to the financial woes of the owner.
Other than physical damage in the form of broken components from an impact, the simple jarring of a watch has the ability to shift the hairspring up or down between the banking pins, allowing it to rub against the balance arms or the balance cock, ruining the accuracy.