A mechanical watch does not keep the same time in every orientation. Lay it flat and it runs at one rate. Stand it on its side and it runs at another, and on a good many watches the difference between the two is larger than the rate itself.
No fault is involved. Gravity acts on a balance wheel that weighs something, and every mechanical watch ever made shows this to some degree.
The positions, and why they differ
Bench testing conventionally uses six static positions. Flat on its back or its face, which are dial up and dial down, and standing on each of its four edges, which are crown up, crown down, crown left and crown right.
Ratekeeper records a different six, and the difference is deliberate. It keeps five of the bench positions and replaces crown right with on the wrist, because a worn watch is where an owner’s rate actually comes from and no bench position stands in for it.
The meaningful split is between flat and upright. Lying flat, the balance wheel turns on the end of its pivot. Standing on an edge, it rests against the side of its bearings instead, and the friction it meets is different. That difference shows up as a difference in rate.
Which direction it goes is particular to the watch. There are rules of thumb in circulation and they are only that, because the answer depends on how the individual watch was adjusted. The reliable way to know what yours does is to measure it, which takes a few nights and no equipment.
Why it matters more than it sounds
A watch that is worn during waking hours and set down at night spends roughly a third of every day in one position of your choosing.
That is a large enough share to move the daily average noticeably. If your watch runs fast lying dial up and you leave it dial up every night, you have been adding to the gain every night without knowing it. Left on its side instead, the same watch might come out closer to true, having done nothing differently in the sixteen hours it spent on your wrist.
This is the reason two people can own the same model and honestly report different rates.
Finding out what yours does
The method is the ordinary one, repeated with one thing changed each time.
- Set the watch against a reference you trust, and note the moment.
- Wear it as usual through the day, and rest it in one chosen position each night.
- After two or three nights, read it and work out the rate for that stretch.
- Set it again and repeat with a different resting position.
Two or three nights per position is the minimum worth bothering with, for the same reason any rate wants a span of days behind it. A single night divides your own reading error by a third of a day and turns a small misjudgment into a large-looking number.
Keep everything else as similar as you can while you do this. Wear it the same amount, keep it wound the same way, and change only where it sleeps.
The one adjustment you can make
Once you know which position runs your watch fast and which runs it slow, you have a lever, and it is the only lever an owner has that involves no risk at all.
If the watch gains over a day, rest it at night in whichever position runs slowest. The two work against each other and the daily average comes in closer to true. If it loses, do the reverse.
Nothing about this touches the movement. You are choosing which of the watch’s own existing behaviours to spend the night in, and you can undo it by putting the watch down differently tomorrow. It will not turn a wild watch into a precise one, and on a watch with a wide spread between positions it can be worth several seconds a day.
What it will not do
Positional compensation moves the average. It does not narrow the spread, and the spread is the thing a watchmaker can actually work on, using methods that live entirely on the bench and have nothing to do with where a watch spends the night.
It also cannot compensate for something changing. If a watch that behaved predictably across positions stops doing so, that is a change in the watch rather than a problem with your nightstand, and no arrangement of the watch on a shelf is the answer to it.
Where this leads
If you want the whole picture rather than the overnight part, how to measure watch accuracy without a timegrapher covers the general method, and why your watch runs fast or slow covers the other things moving your number besides gravity.