Precision: repeatability without variation
Accuracy: ability to hit the desired target
All time-keeping devices I've seen or read about operate either on a "pulse" or "oscillator" whether they're electronic, electro-mechanical or completely mechanical. Their ability to measure passage of time ("time-keeping") relies on both ability to repeat the pulsing or oscillation with as little variation as possible, and the (mean or average) frequency of the oscillator being set as closely as possible to the desired frequency. The amount of variation in its repeated oscillations is a measure of its precision. The mean (average) frequency of the oscillator as compared to it's proper value, is its accuracy.
It's not possible to set accuracy to an extremely tight tolerance unless there is extremely high precision. This is a problem encountered by those trying to tweak watches to extreme accuracy (exceeding that of certified chronometers) that have "off the shelf" workhorse mechanical movements. It is an attempt to achieve an accuracy that exceeds the movement's (current) precision, especially when worn daily which varies the movement's orientation and to some extent, its temperature. Gravity and the movement's position relative to it has a measurable effect on mechanical movements, as does motion of the movement when in use. Temperature affects both mechanical and quartz movements.
In horology, a timepiece with high precision is described as being "well adjusted." A very low "drift" (gain or loss over time) compared to a time standard such as that provided by the U.S. Naval Observatory (USNO) or the Nat'l Institute of Science and Technology (NIST; formerly the Nat'l Bureau of Standards, NBS) is characterized as being "well regulated." Before regulating a watch, it must be adjusted for sufficient precision that allows it to be regulated to the desired accuracy. A watch's precision (how well it's adjusted) ultimately limits its accuracy (how well it can be regulated), no matter how much its regulation is "tweaked."
While many things can be done in most mechanical watches to adjust them, how well the movement is designed, especially the balance wheel, escapement, and mainspring can make adjustment easy or very difficult, and it establishes a baseline precision from which adjustment is made (if desired) to improve it. Three major environmental factors affect their adjustment:
- orientation to gravity,
- motion (including vibration), and
Gravity and Watch Orientation:
This causes minor differences in escapement operation and the bearing of shafts on pivot points that varies the friction in them. There are six basic positions in which a watch can be tested for the effect of gravity on its beat rate: dial up, dial down, crown up, crown down, crown left, and crown right. Adjusting a watch for precision in all these positions is very time consuming and very expensive. It requires tweaking the exact position of staff and shaft pivots . . . and can require tweaking the balance wheel hairspring (usually done to reduce isochronal error).
Motion of the person carrying or wearing the watch, particularly if it rotates the movement, or vibrates it, combined with its orientation to the motion can affect the how many degrees the balance wheel rotates. If there is isochronal error related to balance wheel rotation, the balance wheel period will vary, and timekeeping will vary accordingly.
Changes cause expansion or contraction of all the parts, changing their dimensions, particularly the balance wheel diameter and the hairspring. Materials such as bimetallics and designs that inherently compensate for temperature changes to maintain the same tension on springs, and the rotational inertia of rotating parts help reduce temperature effects. A temperature change and how tightly the mainspring is wound will shift a watch's regulation. How much it shifts is a matter of how much the temperature changes, and how well the design of the balance, hairspring, escapement and mainspring compensate for temperature change, and the mainspring maintaining constant tension (force) as it unwinds from fully wound to fully unwound.
Stringent chronometer standards test timepieces for timekeeping accuracy (its regulation) while inducing conditions that can cause variation at different temperatures and in different orientations called positions (its adjustment). As much hype as the Swiss COSC creates about its chronometer certifications, it is the least stringent of the three major standards that have existed. Furthermore, it's done on the bare watch movement, without any complications, and without its auto-wind rotor (if it's an auto-wind movement). Final assembly of the movement, and assembly into the watch case is done afterward, followed by shipping to the retailer. The U.S. Railroad Standard established at the end of the 19th Century required greater accuracy, and it was performed on the completely assembled watch, just as it would be used. The most stringent was the British Kew Observatory standard established for naval and maritime navigation chronometers. Its complete test required 45 days! If still done (which I doubt), it's now performed by the British NPL (National Physics Laboratory). Kew chronometer testing was also performed on completely assembled chronometers.
Given the materials and modern designs that limit, prevent or compensate for variation of a mechanical movement's regulation, the most important remains positional adjustment. Inside older, higher end watches, it's not uncommon to find the number of positional adjustments made when the watch was manufactured. Typical is either three positions, five positions or "unadjusted" (zero positions; all testing was likely dial up). Rarely is a wrist or pocket watch adjusted for all six positions. The most common position omitted for a wristwatch is crown right (or "12" up). Which is omitted for a pocket watch usually depends on whether it has an open face (no cover over the crystal) or is a "hunter" or "field" style with a hinged metal cover over the crystal.
Most modern Japanese and Swiss movements found in the mid-range to high-range watches are unadjusted workhorse Miyota, Seiko and ETA movements with 17 or more jewels. Their design, particularly the balance, escapement, hairspring and mainspring, have made adjusting movements for this market range unnecessary. Factory regulation can get their accuracy well within about 20 seconds daily gain/loss, the common factory specification. It's almost always much better than that (approx. +/- 10 sec./day). Their variation in daily gain/loss rate (precision) is usually much less than +/- 5 seconds per day.
These watches can quite often be regulated later by a watchmaker experienced with mechanical movements and knows what he's doing, to within 10 seconds gain or loss per day without any adjustment. That's a little over a minute per week, and it assumes the movement in good mechanical working condition (good lubrication, no corrosion, etc.). He'll keep it a few days although he won't spend much bench time working on it. The calendar time is needed to let it run a day so between tweaking its regulation a couple times, and then verify after another day of running that his tweaking hasn't gone too far in either direction. Some of how well he can tweak it depends on the balance regulator and how finely it allows moving it. Some have a screw for fine tuning its regulation; others do not.
Some exceptions in the high-range might be those destined to undergo COSC chronometer certification. COSC is expensive enough to push these exceptions into the upper end of high range pricing. It depends on the watchmaker whether or not the movement is adjusted before it's submitted to COSC for testing. Adjustment is most typically reserved for the luxury and prestige watch movements. It's very time consuming and quite costly compared to regulating a movement. Many of the movements from the most widely recognized of these brands are not only adjusted to three or five positions, they're COSC certified (e.g. Rolex and Omega).
Attempting to achieve +/-4 second regulation or better with any mechanical watch movement, no matter how well designed and adjusted, even the COSC certified prestige and luxury watch movements, is nearly impossible if the watch is worn daily. Such regulation accuracy exceeds that of the British Kew Observatory Standard! In my very humble opinion, achieving better than +/-20 seconds in a 24 hour period, typical of the factory specifications for unadjusted movements in the inexpensive, basic lines of mechanical watches sold for less than $200, with many under $100, is very good, and quite a few are much better than that out of the box. That unadjusted movement mid-range and high-range watches are typically +/- 10 second daily gain/loss out of the box is excellent.