Answer
How accurate is an automatic watch?
Short answer
A healthy modern automatic keeps time to roughly minus 10 to plus 20 seconds a day, and a chronometer-certified movement is tested to a mean daily rate of minus 4 to plus 6 seconds a day. Judge yours on a seven-day average rather than a single morning, because position, temperature and winding state move the rate every day.
On this page (5 sections)
An automatic watch is a mechanical watch, so it keeps time to seconds a day rather than seconds a month. The realistic range for a modern automatic in good health is roughly minus 10 to plus 20 seconds a day. A chronometer-certified movement is tested to a mean daily rate of minus 4 to plus 6 seconds a day, and a well-adjusted uncertified movement frequently lands in the same territory once it has settled.
Those numbers only mean something as averages. What follows is how to measure the right thing, what the published tolerances actually are, and how to tell an ordinary rate from a symptom.
Accuracy means mean daily rate, not yesterday
A mechanical watch does not have one rate. It has a different rate in every position, at every temperature and at every state of wind. So every standard, and every sensible owner, measures a mean daily rate: total gain or loss divided by the number of days, over at least a week of normal wear.
This is why the commonest complaint is not measurable. "It gained eight seconds yesterday" mixes one day of genuine variation with your own reading error at both ends of the measurement. A watch that gains four seconds on Monday and loses two on Tuesday is running at plus 1 second a day, which is excellent.
The underlying error is small even when it feels large. At 28,800 vibrations per hour, which is 4 Hz or eight beats per second, the balance makes about 691,000 beats a day. Gaining ten seconds means being wrong by about 80 of them, or roughly one part in 8,600.
What the published tolerances are
| Tier | Typical published or observed rate | Notes |
|---|---|---|
| Budget Japanese automatic | roughly minus 35 to plus 45 seconds a day | A wide published tolerance, though real examples usually do far better |
| Standard-grade Swiss automatic | roughly plus or minus 12 seconds a day | Typically adjusted in two positions at the factory |
| Upgraded or top-grade movement | roughly plus or minus 4 to 7 seconds a day | More positions adjusted, better regulating hardware |
| COSC chronometer | minus 4 to plus 6 seconds a day | Mean daily rate, tested on the uncased movement |
| METAS Master Chronometer | 0 to plus 5 seconds a day | Tested on the finished, cased watch after 15,000 gauss exposure |
Note the asymmetry in the certified figures. Standards allow a watch to gain more than it loses, because a movement running slightly fast is generally the sign of an oscillator with energy to spare, while one running slow is often one losing energy to friction.
Why the rate moves from day to day
Five things account for almost all of the variation an owner sees.
- Position. Dial up, dial down and each crown-vertical position give different rates, sometimes several seconds a day apart. Your daily rate is a weighted average of however you happened to move that day.
- State of wind. Rate drifts as the mainspring unwinds. A sedentary day leaves an automatic underwound, and the last hours of the power reserve are the least accurate hours it has.
- Temperature. Metal balances and hairsprings change dimension with heat. A summer week on the wrist and a winter week on a cold nightstand are not the same test.
- Magnetism. The single most common cause of a sudden, dramatic gain. It is also the cheapest to fix.
- Condition. Aging oil and worn pivots cost the balance energy, and a movement with low amplitude keeps worse and less consistent time.
Judging your own watch
Wear it for a week, compare it each morning against a phone or computer clock that syncs over the internet, and divide the total deviation by seven. Then check that average against three things: the tolerance the manufacturer publishes for the caliber, the size of the day-to-day spread, and whether the figure is stable from one week to the next. Our guide to timing your own watch sets out the method in full, and the accuracy calculator does the arithmetic.
If the weekly average sits inside the published tolerance, the watch is doing its job, however unglamorous the number looks. If it sits outside but is consistent, ask a watchmaker for a regulation rather than assuming a fault.
When the number is telling you something
Certain patterns are worth acting on rather than tolerating.
- A sudden change of minutes a day. Suspect magnetism first. Demagnetizing takes seconds and typically costs nothing at a watchmaker who is already looking at the watch.
- A rate that is fine at full wind and poor by the second day. A torque or mainspring issue, not a regulation issue.
- A steadily worsening loss over months. Usually degraded lubricant, and the cue for a service conversation.
- A watch that runs well worn but stops overnight. Low amplitude that survives motion but not stillness. That is a bench job.
What is not realistic is quartz-like steadiness. A mechanical rate depends on gravity, temperature, friction and the state of a metal spring, and it will move by a second or two a day for reasons nobody can control. If you need to be within a few seconds a month, quartz is the honest answer. If you want a machine that is good enough to trust and interesting enough to keep for fifty years, an automatic within ten seconds a day is a success.
Common follow-up questions
Does an expensive automatic keep better time than a cheap one?
Often, but not reliably. Price buys finishing, materials and brand as much as adjustment. A watch costing around $400 that has been regulated well can beat one costing $6,000 that left the factory set two positions off. What price does buy consistently is a tighter published tolerance and a movement that holds its rate longer between services.
Will my automatic get more accurate as it breaks in?
Slightly, and mostly in the first few weeks as the lubricants distribute and the mainspring settles. Do not expect a watch that runs plus 25 seconds a day out of the box to drift back to zero on its own. If it is still far outside the maker's published tolerance after a month of normal wear, that is a warranty conversation.
Is it normal for my watch to be accurate one week and worse the next?
Some week-to-week movement is normal, especially if one week was active and the next was sedentary, or if the seasons changed. A swing of a second or two in the weekly average is ordinary. A swing of ten seconds a day between weeks points at inconsistent winding, magnetism or a mechanical problem.
Do automatics keep better time than hand-wound watches?
Not inherently. The oscillator does the timekeeping and it is the same in both. Automatics have one practical advantage: worn daily they sit near full wind, where the rate is most stable. A hand-wound watch you wind irregularly spends more time at the weak end of the mainspring, where the rate drifts.
Sources and further reading
- ISO 3159, Timekeeping instruments: wrist-chronometers with spring balance oscillator, for the multi-position, multi-temperature test protocol and the mean daily rate tolerance.
- Contrôle Officiel Suisse des Chronomètres (COSC), published certification criteria for mechanical wristwatch movements.
- METAS, Master Chronometer certification documentation, for the 0 to plus 5 seconds a day tolerance measured on the cased watch.
- Movement manufacturers' technical data sheets (ETA, Sellita, Seiko, Citizen), which publish rate tolerances by grade and the nominal frequency of each caliber.
- Witschi Electronic, timing machine operating documentation, for rate, amplitude and beat error measurement on mechanical movements.
Last reviewed 4 September 2026. Spotted an error? Tell us and we will fix it in public.