A timing machine listens to a watch tick and reports what it hears over a few seconds. It is a fine instrument and it costs money, needs the watch held in a cradle, and tells you about those few seconds rather than about your week.

The alternative is older than the machine and needs nothing you do not already own. Set the watch against a clock you trust. Wait. Look again. The difference, divided by the time that passed, is your rate in seconds per day.

The method

Set the watch to a reference you trust. Not the clock on your oven. Your phone is close, though the second it displays is not necessarily the true one, since phones set themselves from a variety of sources and the display lags the moment slightly. A network time reference is the honest version and is what an app can reach directly.

Bring the second hand round to twelve and let the crown out so the movement stops, then push it in at the exact moment your reference passes the minute. This is called hacking and most modern mechanical watches do it. If yours does not, do not fight it. Leave the watch running, note how many seconds it already differs from your reference, and treat that as the starting offset rather than trying to zero it. The arithmetic works from any two known offsets, not only from zero.

Note the moment precisely. Write down the date and time to the minute. A reading you cannot place in time is not a reading.

Wait. A day or two is the useful minimum. Longer is better and cheaper, since every additional day divides your own imprecision further. A week is a comfortable interval and gives a figure you can lean on.

Read it against the same reference. Wait for your reference clock to cross an exact minute, look at the watch, and record how far ahead or behind it is in seconds. Reading the watch as it crosses twelve is easier than reading it at an arbitrary moment, so it is usually worth waiting a few seconds for that.

Divide. Seconds of drift, divided by days elapsed. The calculator does that step, or a pocket calculator will.

What to hold steady

The number you get describes the watch under the conditions it lived through, so the conditions are part of the measurement.

  • Wear it as you normally do, or do not wear it at all, but do not do half of each and expect a clean answer. A watch worn for two days and left in a drawer for two more has given you an average of two different situations.
  • Keep the resting position consistent overnight. Gravity moves the rate, often by more than the rate itself, so the position a watch spends its nights in is quietly a large part of the answer. Resting position and overnight accuracy covers how much difference it makes.
  • Keep it wound. An automatic on a wrist stays near a full mainspring. One that runs down partway through the interval spends the tail of that interval behaving differently, and you will be measuring the running down as much as the watch. Power reserve, winding, and rate covers how much wearing actually winds one.

You cannot control all of this perfectly, and you do not need to. You need to know which of them you did not control, so you know what the number is describing.

The mistakes that spoil a reading

Measuring over too short a span. This is the common one. If you misjudge either reading by a second, and there are only twelve hours between them, that one second becomes two seconds a day of error in the answer. Over a week it becomes a seventh of a second.

Confirming the wrong minute. Watching the second hand hit twelve tells you where the seconds are, not which minute you are in. Read the minute hand carefully, especially near the hour, where a minute hand sitting just short of the mark is easy to round the wrong way.

Comparing against a drifting reference. If you set the watch against one clock and read it against another, you have measured the difference between two clocks and one of them is not your watch.

Changing something mid-interval. Setting the watch, winding a manual movement fully, taking it through an airport, or leaving it somewhere cold all belong to the interval they happen in. If something notable happened, start again rather than explaining it away.

One reading is an anecdote

A single rate tells you about one interval. What tells you about a watch is a series of them, because a watch that reads +4 three times running is a different proposition from a watch that reads +4 once and +11 the next time.

That is the case for keeping the readings rather than the arithmetic. Once you have half a dozen, patterns appear that no single figure contains, and a change in the pattern later is the genuinely useful signal.

Where the app fits

Ratekeeper is this method with the fiddly parts removed. It reads network time directly rather than the phone’s display, takes the reading from a tap as the second hand crosses twelve, does the division, groups readings into runs so that resetting the watch does not corrupt the history, and records which position the watch rested in so the positional part stops being guesswork.

None of that is a secret and none of it is magic. It is the same arithmetic, kept for you. It is free for your first watch.