Watch accuracy calculator

How to measure honestly

Set the watch against a reference you trust: a phone clock, which is synchronized to network time, or an online atomic clock. Note the exact offset. Wear the watch normally for seven days, then read the offset again at the same time of day. The difference divided by seven is your rate.

Three habits separate a measurement from a guess. Read the offset to the second, by watching the seconds hand cross the twelve and noting what the reference says at that instant. Record the starting offset, because the tool wants drift rather than current error: a watch set 4 seconds fast that now reads 35 seconds fast has drifted 31. And enter the window as you measured it, in the half-day steps the input allows, so an evening reading on day seven is 7.5 days.

The length of the window matters more than people expect, because your reading error is divided by it.

One day is not a measurement. A mechanical watch's rate varies with how wound it is, which positions it spent the night in, and how much you moved. A week averages those out. If you want to go further, note the position it rests in overnight, then repeat the week with a different resting position: the difference between the two is positional error, and it is often the cheapest accuracy gain available to you.

A worked example, one week

You set the watch to an atomic clock at 8:00 on a Monday morning, wear it normally, and at 8:00 the following Monday it reads 8:00:35. Enter 7 days and a drift of plus 35 seconds.

The conclusion is that this watch needs nothing done to it, beyond resetting to the second every week or two if that matters to you. Sloppy inputs move it less than you would fear: reading at 8pm makes the window 7.5 days and the rate 4.7, and forgetting a 4 second head start makes it 4.4. Both land in the same band, which is the point of a week-long window.

What the standards actually require

COSC certification applies the test regime in ISO 3159 to an uncased movement: 15 days, 5 positions, 3 temperatures at about 8, 23 and 38 degrees Celsius (46, 73 and 100 Fahrenheit). The headline requirement is a mean daily rate of minus 4 to plus 6 seconds a day, but seven criteria are computed and the other six are about consistency, among them a cap of 2 seconds a day on the mean variation in rates across positions. A movement can post an excellent average and still fail for being erratic.

METAS Master Chronometer certification starts from a COSC-certified movement and tests the cased watch, requiring 0 to plus 5 seconds a day after exposure to a 15,000 gauss magnetic field. Rolex publishes its own cased-watch figure of minus 2 to plus 2. None of these describes what your watch will do on your wrist, because none of the tests involves a watch being worn. The comparison of COSC, METAS and brand standards sets them side by side.

A good non-certified automatic in daily wear typically lands between minus 10 and plus 20 seconds a day. Entry level movements are often specified more loosely still, and a watch inside its own manufacturer's specification is not faulty even if it annoys you.

What the verdict band does not mean

The tool judges the size of the rate, not its sign, so minus 6 seconds a day is reported as COSC-grade when COSC itself would reject it: the standard allows only 4 seconds slow against 6 seconds fast, because a movement that gains can be regulated down more predictably than one that loses. Read the sign yourself.

The bands are also mechanical bands. Feed in a quartz watch and 15 seconds a month arrives as 0.5 seconds a day and is called chronometer territory, when it is merely ordinary for the technology. A thermocompensated movement specified at around 10 seconds a year is 0.03 seconds a day, which takes months rather than a week to separate from noise. See quartz accuracy.

Finally, an average hides its own spread. A watch that ran plus 25 one week and minus 15 the next averages plus 5 and looks immaculate here. Consistency is what a chronometer certificate actually buys, and it is what a timing machine reading shows in a way a week of wear cannot.

When a rate is telling you something is wrong

A watch that suddenly starts gaining a large amount, especially more than about 30 seconds a day when it used to be fine, is usually magnetized. That is a five minute fix at a watchmaker and costs little or nothing, and our guide to magnetism explains why a hairspring with its coils clinging together runs fast. A watch that starts losing time, runs unevenly, or stops before its stated power reserve is more likely to be low on amplitude: old oil, wear, or a tired mainspring. That is a service, not a regulation.

Some causes are not faults. An automatic worn two days a week never reaches a full mainspring, and a movement running on the last of its reserve loses time. A watch left crown up overnight can differ by several seconds a day from the same watch left dial up, and a cold week differs from a hot one, which is exactly why the standards specify three temperatures.

What to do with the number

The rate you just calculated should drive one of four decisions.

1. Within about 10 seconds a day and steady: do nothing. Try a different overnight resting position for a week if you want the last few seconds for free.
2. Consistently 15 to 30 seconds a day: ask for regulation, roughly $60 to $150 in the US before sales tax, and often folded into other bench work. It adjusts the rate, not the health of the movement. See whether a watch can be regulated.
3. A sudden jump beyond 30 seconds a day: have it demagnetized first. It is quick and cheap, and if the rate returns you have saved yourself a service.
4. Losing time, with a short power reserve: that is a full service, indicatively $250 to $500 at an independent watchmaker and more at a brand center, before sales tax.

Our guide to timing your own watch covers the procedure in more detail, and the guide to accuracy and chronometer standards covers where every threshold on this page comes from.