Does cold weather make a watch run fast or slow?

Cold makes a quartz watch run slow, not fast. The tuning fork in an ordinary analog quartz watch runs fastest at about 25 degrees Celsius, and every degree away from that point, hotter or colder, slows it by an amount that grows with the square of the distance. One of the answers that ranks for this question says quartz gains in the cold. The crystal makers' data sheets say it loses.

Mechanical watches are where cold can make a watch gain. A plain steel hairspring stiffens as it cools and the watch speeds up, which is where the "fast in the cold" idea comes from. Modern hairspring alloys cancel most of that and thickening oil pushes back, so for a current mechanical watch the honest answer is a tolerance, not a direction.

Why a quartz watch runs slow in the cold

A quartz watch counts the vibrations of a tuning fork at 32,768 Hz, as our page on how accurate a quartz watch is explains. Raltron's R38 data sheet, for a 3 by 8 mm tuning fork crystal, gives its temperature behavior as Δf/f = K(T₀ − T)², with the turnover temperature T₀ at 25 ±5 °C and K at -0.034 ±0.006 ppm/°C².

The square and the minus sign settle the question. Squaring makes the temperature difference positive whether the watch is 10 degrees colder or 10 degrees warmer, and the negative K turns that into a lower frequency either way. Fewer vibrations per real second means the hands fall behind. No temperature makes an ordinary crystal run faster than it does at turnover.

Other makers agree. Diodes Incorporated prints -0.03 ±0.01 ppm/°C² and the same 25 ±5 °C turnover for its XK16 series, with a curve that falls away on both sides. ECS Inc. calls 0.04 ppm/°C² "a common parabolic coefficient" for a 32.768 kHz tuning fork.

The arithmetic, worked

One part per million is 0.0864 seconds a day. With Raltron's typical K and a turnover at exactly 25 °C:

At freezing the crystal loses about 21 ppm, as much as the whole factory tolerance of a standard part: ECS gives ±20 ppm as "gaining or losing 1.7 seconds of time each day". Across the K values the makers print, 0.02 to 0.04, the 0 °C loss falls between about 1.1 and 2.2 seconds a day.

The loss is charged by the hour. A watch left eight hours in a car at -10 °C loses a third of 3.6 seconds, about 1.2 seconds, then returns to its normal rate. Near the top the curve is gentle: between 20 and 30 °C the loss stays under a tenth of a second a day.

Why the "fast in the cold" answer keeps circulating

Because it is true of a different watch. Seiko's patent on a temperature-compensated balance (US 9,188,956) explains that steel's Young's modulus has a negative temperature coefficient, so warming softens the hairspring, while the balance wheel expands and gains inertia. The result is that "the timepiece is fast at the low temperature and slow at the high temperature". Carry that rule across to quartz and it gives the wrong answer.

The factory trim adds to the confusion. A standard crystal may leave the factory up to 20 ppm fast. A watch trimmed 10 ppm fast still reads fast on a cool day, only less so; its owner has noticed the trim, not the cold.

Mechanical watches: two effects pulling against each other

The hairspring. Steel stiffens as it cools, so the balance swings back faster and the watch gains. Wikipedia's history of the balance spring describes the compensation balances built to offset this, and then Charles-Édouard Guillaume's Elinvar, a nickel-steel alloy whose modulus of elasticity is "essentially unaffected by temperature". The Seiko Museum dates Elinvar to 1913 and Nivarox, a harder Elinvar-based alloy, to 1933. Modern hairsprings use alloys of this family or silicon, so the effect is small rather than gone.

The oil. Watch oil thickens as it cools. Citizen's lubricant patent (US 6,858,567) gives a low-viscosity oil it used in the rotor section as 16 cSt at 50 °C and 840 cSt at -20 °C, more than fifty times thicker. Thicker oil means more friction, friction takes energy from the balance, and its amplitude falls. How far the rate follows the amplitude, and in which direction, depends on the movement's isochronism and adjustment, so no one can name a direction for every watch.

In deep cold the oil wins outright. The same patent notes that watches have a problem at low temperatures: "operation failure takes place when the temperature becomes lower than -10° C". Webchronos, describing Sinn's watches, says ordinary oil becomes viscous at around -25 °C, "making it difficult to maintain the accuracy of the watch", and that Sinn's own oil holds DIN-specified accuracy down to -45 °C. Moebius rates its thin 9030 oil at -40 to +60 °C.

What COSC allows in a cold week

For a modern mechanical watch the useful published number is a ceiling. COSC says movements are "successively placed in chambers at 8°, 23°, then 38°", and applies the seven criteria of ISO 3159, one of which is rate variation with temperature: ±0.60 seconds a day per degree, as our page on what COSC certified means lists.

So a certified chronometer running at zero at 23 °C may run anywhere from -9 to +9 seconds a day at 8 °C and still pass, because 15 degrees times 0.6 is 9. Most do far better, but uncertified movements are held to no figure at all.

The quartz chronometer limit, from Wikipedia's account of the COSC criteria, is beyond a bare crystal, so those movements measure their own temperature and correct for it. And the mechanical allowance is ten times what cold does to quartz, one reason a mechanical watch timed in January and again in July can give two different, equally healthy answers.

Why the wrist matters more than the weather

A worn watch is not at air temperature. Wikipedia's quartz clock article describes a well-designed case as "an expedient crystal oven that uses the stable temperature of the human body" to hold the crystal near its best range. If the case sits at 32 °C while worn, the loss is 0.14 seconds a day, about 4 seconds a month. A quartz watch in a drawer at 10 °C loses five times as much.

The curve is symmetrical, so a nightstand at 18 °C costs a quartz watch the same 0.14 seconds a day as the wrist. Off the wrist only starts to matter below about 15 °C.

For a mechanical watch, a night off the wrist changes temperature, position and state of wind at once, and position alone commonly moves the rate more than temperature does, as our pages on a watch that runs fast or runs slow explain. A different overnight result is not, on its own, evidence about cold.

How to tell whether cold is your problem

For quartz, the effect is too small to see in a day. Set the watch against an internet-synced clock and read it after 30 days, as our home accuracy test describes, then compare the loss with the 30-day column above.

For a mechanical watch, change one thing at a time. Wind fully, rest it overnight in one fixed position at room temperature, and read it against a synced clock in the morning. Repeat with the same position and wind somewhere cooler. The difference between the nights is the watch's thermal behavior. Our guide to timing your own watch covers reading to the second, and the accuracy calculator does the arithmetic.

A mechanical watch that slows sharply or stops in real cold has an oil problem, which a service addresses. Fog under the crystal after coming in from the cold is a seal problem, not a temperature one, as our page on fog under the crystal explains.

When this answer stops applying

• The turnover is not exactly 25 °C. Both makers allow ±5 °C. A crystal turning over at 20 °C loses about 1.2 seconds a day at freezing, one at 30 °C about 2.6. The direction never changes.
• The watch is thermocompensated. It cancels most of the curve, so the table does not apply.
• The mechanical watch has a plain steel hairspring. It follows the steel rule and gains in the cold.