Answer
How do I test my watch's accuracy at home?
Short answer
Set the watch against an internet-synced clock, then record the total deviation at the same time each morning for seven days and divide by seven. That average is your mean daily rate. A phone timing app or a timegrapher adds amplitude and beat error, which diagnose the cause, but only the wrist test tells you how the watch performs in your life.
On this page (4 sections)
Almost every complaint about watch accuracy comes from bad measurement. Somebody glances at a phone one morning, sees the watch is eleven seconds out, and concludes it is running fast. What they have measured is an unknown number of days of drift, plus an unknown starting error, plus their own reaction time, divided by nothing.
Testing a watch properly at home needs no equipment and seven days of mild discipline. Everything beyond that, apps and timing machines, answers a different question: not how well the watch keeps time, but why.
The seven-day wrist test
- Pick a reference. A phone or computer clock set to update automatically, or a website showing official time.
- Start at full wind. Wind a manual watch fully; wear an automatic long enough to be fully wound. Starting low measures the weak end of the mainspring rather than the watch.
- Set it exactly. If the movement hacks, pull the crown, set the hands slightly ahead, and push the crown in as the reference passes that time. If it does not hack, set it as close as you can and write down the offset as your day-zero figure instead of pretending it is zero.
- Read it once a day, at the same time. Record the total deviation from the reference in seconds, not the change since yesterday. A reading taken over 23 hours compared with one taken over 25 is noise.
- Note the context. Hours worn, active or sedentary day, and the overnight resting position. This column is what turns a number into an explanation.
- Divide by seven. Plus 31 seconds after a week is a mean daily rate of about plus 4.4 seconds a day.
Wear and store the watch exactly as you normally would. The point of the test is your life, not a laboratory. The accuracy calculator will convert a set of daily readings into a mean daily rate and a projected monthly figure.
The positional test
If the wrist test gives an inconsistent answer, this is the follow-up. Wind the watch fully, leave it in one position for 24 hours, record the deviation, then reset and repeat for each of six positions: dial up, dial down, crown up, crown down, crown left and crown right.
What you want is not any single figure but the delta, the spread between the fastest and slowest position. A watch running plus 2 in every position is beautifully adjusted. A watch running minus 8 dial up and plus 14 crown down will give you a different weekly average depending on what you did that week. Delta is also the number that tells you a regulator arm will not solve the problem.
A useful side effect: once you know which position loses, you can rest the watch that way overnight and pull the daily average toward zero without touching the movement.
Phone timing apps
Several apps listen to the escapement through the phone microphone and report a rate in seconds a day within a minute or two. They are genuinely useful, with limits worth knowing.
- They need quiet and contact. Rest the watch directly on the phone near the microphone, in a silent room. Room noise, air conditioning and traffic will produce nonsense.
- Rate is the reliable output. Amplitude readings from a phone microphone are far less trustworthy than from a proper machine, because a phone mic is not designed to hear the three-part escapement sound cleanly.
- You must tell it the right frequency. A movement at 28,800 vibrations per hour is 4 Hz, or eight beats per second. Set the app to 21,600 vph by mistake and every number it gives you is wrong. Our beat rate converter translates between the units.
- It is a snapshot. One position, one moment, one state of wind, which is not the same thing as a mean daily rate.
What a timegrapher adds
A timegrapher clamps a piezoelectric sensor against the case and times the escapement far more precisely than you can time hands against a clock. Entry-level machines cost roughly $150 to $250 before sales tax, with prices outside the United States differing by local tax and import duty, and they report three numbers.
| Reading | Healthy range at full wind | What it tells you |
|---|---|---|
| Rate, seconds per day | Within a few seconds of target, in that position | Whether regulation is needed |
| Amplitude, degrees | Roughly 270 to 310 dial up | Movement health: below about 250 suggests a service |
| Beat error, milliseconds | Under about 0.5 ms is good, under 1.0 ms acceptable | Whether tick and tock are even |
Use each tool for what it is good at. Diagnose with a machine, judge with the seven-day test, and if the two disagree, the wrist is the one you have to live with. Our guide to timing your own watch goes deeper on reading the trace, and the timegrapher reading explainer interprets a set of numbers you already have.
Common follow-up questions
What should I use as a reference clock?
A phone or computer clock set automatically, which synchronizes over the internet, or a website that displays official time. Do not use another wristwatch, a microwave, a car dashboard or a wall clock. Measuring against an unverified clock measures two unknowns at once, and the reference is often the worse of the two.
Should I test the watch on my wrist or leave it on a table?
On your wrist, worn exactly as normal, because that is the question you are asking. A watch left on a table for a week measures one position at a declining state of wind, which is a laboratory result nobody lives in. Table testing is useful only as a deliberate single-position test.
How accurate can my own measurement be?
Your reaction time at each end is worth a second or so, and reading a seconds hand against a digital display adds a little more. Over seven days that error is divided by seven, so a careful home test resolves a mean daily rate to well under a second a day, which is finer than the differences most owners are arguing about.
Can I test a watch that has no seconds hand?
Yes, but the resolution is poorer. Note the exact instant the minute hand crosses a marker at the start and compare it with the same crossing a week later, using a magnifier if the dial allows. Expect to resolve a few seconds rather than one, and run the test for two weeks to compensate.
Sources and further reading
- ISO 3159, Timekeeping instruments: wrist-chronometers with spring balance oscillator, for the multi-position, multi-day protocol that home testing imitates.
- Contrôle Officiel Suisse des Chronomètres (COSC), published criteria including mean daily rate and difference between positions.
- Witschi Electronic, timing machine operating documentation, for rate, amplitude, beat error and the role of lift angle in amplitude calculation.
- Movement manufacturers' technical data sheets (ETA, Sellita, Seiko, Citizen), which publish lift angle, nominal frequency and rate tolerance by caliber.
- Donald de Carle, Practical Watch Repairing, for the interpretation of positional rate differences and regulation.
Last reviewed 4 September 2026. Spotted an error? Tell us and we will fix it in public.