How Accurate Should a Mechanical Watch Be?

A healthy mechanical watch typically runs within about −10 to +20 seconds per day, and many modern movements do noticeably better. A COSC-certified chronometer must keep −4/+6 seconds per day, Omega's METAS-certified watches are held to 0/+5, and Rolex promises ±2. If your watch stays inside its manufacturer's published range and gains or loses a consistent amount each day, it is working as designed.

What counts as normal accuracy

“Accuracy” for a mechanical watch is usually expressed as a daily rate: how many seconds the watch gains (+) or loses (−) in 24 hours. Unlike quartz, a mechanical movement keeps time with a balance wheel — a tiny oscillating flywheel — and its rate shifts with position, winding state, and temperature. So there is no single number; there is a range, and the range depends on what grade of movement you own.

Rough tiers, based on manufacturers' published tolerances:

Why do tiers exist at all? Often the higher grades are the same basic movement, adjusted and tested more carefully. An ETA 2824-2 in top grade differs from the standard grade mainly in regulation (fine adjustment of the balance), the quality of the balance spring and shock protection, and how many positions it was adjusted in at the factory. That's also why a well-regulated budget movement can outperform its paper spec: the tolerance describes the worst the manufacturer will ship, not the best the design can do.

Two things matter more than the raw number. First, consistency: a watch that gains 12 seconds every single day is easy to live with (and easy for a watchmaker to regulate); a watch that gains 5 one day and loses 15 the next has a problem even though both numbers look modest. Second, direction: many owners prefer a watch that runs slightly fast, because it's easier to notice and reset than one that quietly falls behind.

Accuracy standards compared: COSC, METAS, Superlative, JIS

Certification standards exist so that “chronometer” on a dial means something testable. Each body publishes its own tolerance and test protocol:

A detail worth knowing: these are average daily rates measured under controlled conditions. On a wrist — with real temperature swings, varied positions overnight, and changing power reserve — a COSC watch can legitimately read slightly outside −4/+6 on a given day and still be perfectly in spec. How brands' published specs compare to what owners actually measure is covered in our guide to expected accuracy by brand.

What affects accuracy day to day

Four factors explain most of the day-to-day variation in a mechanical watch's rate:

Position. Gravity acts on the balance wheel differently depending on orientation. Dial-up, dial-down, and the four crown positions each produce a slightly different rate — that's why certification tests measure multiple positions. You can use this deliberately: if your watch runs a touch fast, resting it crown-down or crown-up overnight often changes the overnight rate enough to compensate. Which position does what varies by movement, so you have to measure your own watch.

Winding state. A mainspring delivers less torque as it winds down, which lowers balance amplitude and usually shifts the rate near the end of the power reserve. This effect is called isochronism error. An automatic worn daily stays near full wind; the same watch left on a desk for two days may time differently in its final hours of reserve.

Temperature. Modern balance springs are made from temperature-compensating alloys, but heat and cold still nudge the rate by a few seconds per day at the extremes — which is why chronometer testing includes three temperatures.

Magnetism. The sneaky one. A magnetized balance spring can make a watch suddenly run dramatically fast — sometimes minutes per day — because the coils stick together and shorten the spring's effective length. Phone cases, laptop speakers, and bag clasps are common culprits. The fix (demagnetization) is quick and cheap. If your watch's rate changed abruptly, start with our guide to why a watch runs fast or slow.

How to measure your watch's real rate

Published specs tell you what the movement should do; only measurement tells you what your watch actually does. The principle is simple: compare the watch against a reliable reference time, wait, compare again, and divide the drift by the elapsed time. Do this over several days — a single 24-hour reading mixes together position and winding effects, while a multi-day average smooths them out. Our step-by-step guide to measuring watch accuracy at home walks through doing this with no equipment at all.

Time is your friend here. Any manual check carries a reading error of a second or two, but that error gets divided by the measurement window: a 2-second reading error over 24 hours muddies the result by 2 s/day, while the same error over four days contributes only 0.5 s/day. Two or three checks spread across several days of normal wear will give you a trustworthy average daily rate. Note the conditions too — worn versus resting, and which position it rested in — because those are exactly the variables that move the number.

The practical challenge is precision: reading a watch dial against a clock by eye introduces a second or so of error at each check, so short measurement windows produce noisy numbers. This is the problem an app like ChronoLog solves — it timestamps each check against atomic time using quick-tap and photo measurement modes, tracks the accuracy history of each watch in a collection, and can even measure rate and beat error acoustically through the phone's microphone (typically within ±2–5 s/day, no extra hardware).

The other route is a timegrapher — a device (or app) that listens to the escapement and reports the instantaneous rate in seconds. It answers in moments what dial-watching answers in days, though it shows a snapshot in one position rather than real on-wrist performance. See what a timegrapher is and how to read one, or the technical deep-dive on building a timegrapher from a phone microphone.

When bad accuracy means a service is due

Accuracy is also a health report. The pattern of the error tells you which problem you likely have:

Gradual slowing over months or years is the classic service signal. Aging lubricants increase friction, balance amplitude falls, and the rate drifts — often slower and more erratic across positions. Most manufacturers recommend a service roughly every five to ten years; a watch that has quietly gone from +3 to −15 s/day since its last service is telling you it's time.

Sudden large gains — a watch that was fine last week and now gains two minutes a day — point to magnetization, not wear. That's a five-minute demagnetizer fix, not a full service.

Sudden erratic behavior after a knock (stopping, wildly inconsistent rate) suggests shock damage to the balance or escapement and deserves a watchmaker's inspection promptly.

A steady rate outside spec — say, a consistent +25 s/day on a movement rated ±12 — often needs only regulation, a minor adjustment of the effective balance spring length, rather than a teardown. Bring a few days of measurements to the watchmaker; a documented daily rate across positions makes their job faster and your conversation concrete.