What regulation means
Picture a watchmaker's bench in the back room of a Geneva manufacture: a Witschi timing machine humming on standby, a felt-lined tray, a half-frame loupe, an expression of practised patience. The watchmaker sets the watch dial-up on the machine's microphone, watches the trace settle for fifteen seconds, then moves a lever no larger than an eyelash. The trace shifts. This is regulation: adjusting a movement's rate so it gains or loses no more than a specified number of seconds per day.
Mechanically, regulation means changing the balance's effective oscillation frequency. In most movements this is done with an index regulator — two curb pins straddling the hairspring's outer coil; sliding them along the spring changes its active length, and a longer active spring means a slower beat. In free-sprung movements there are no curb pins at all: the hairspring breathes untouched, and rate is set by turning tiny calibrated masses or screws on the balance rim, changing its inertia instead. With a timing machine, an experienced hand can measure, correct, and confirm in minutes.
What regulation cannot do is fix positional variation — the difference in rate between orientations. That is the province of adjustment: the slower, deeper work of truing the balance's poise, the pivots, the endshakes, and the hairspring's centring so the rate stays consistent however gravity happens to be pulling. Regulation sets the average; adjustment narrows the spread. The distinction matters because a watch can be perfectly regulated and badly adjusted — right on the desk, wrong on the wrist.
The six positions
Full adjustment means optimising rate across six positions: dial up, dial down, crown up, crown down, crown left, crown right. In each, gravity loads the balance pivots differently — horizontal positions run the pivots on their tips, vertical positions on their sides, with different friction and different rates. A movement adjusted only flat can vary 15 to 30 seconds per day between horizontal and vertical: correct on the nightstand, wrong all day. Free-sprung balances with overcoil or well-formed flat springs — standard practice at Rolex, Patek Philippe, and A. Lange & Söhne — deliver measurably tighter positional spread, because the index pins whose variable contact disturbs ordinary movements simply are not there. "Adjusted to five positions," engraved on a fine American railroad movement a century ago, meant exactly what it means today: someone spent hours making gravity irrelevant.
What do chronometer certifications actually test?
COSC — the Contrôle Officiel Suisse des Chronomètres — tests each uncased movement for fifteen days, in five positions, at 8, 23, and 38°C. The familiar headline figure, a mean daily rate of −4 to +6 seconds, is only one of seven criteria; the others bound rate variation across days, positions, and temperatures, which is the harder discipline. COSC is a meaningful floor, not a ceiling — well-sorted modern movements beat it comfortably. The stricter house standards stack on top: Rolex's Superlative Chronometer certifies the complete cased watch to ±2 seconds per day; Omega's METAS Master Chronometer adds testing during and after exposure to 15,000-gauss magnetic fields, with the cased watch held to 0/+5; Patek Philippe's in-house seal demands −3/+2 for most calibres; Grand Seiko's standard requires −3/+5 over seventeen days in six positions. The Geneva Seal, by contrast, is primarily a finishing and construction certification with a rate requirement attached — it certifies a different virtue. Underneath all of them sits one physical property: isochronism, the oscillator's indifference to amplitude. Every test above is, in the end, a measurement of how well the watchmaker has approximated it.
A timing machine does not measure rate directly — it listens to escapement ticks and computes amplitude from their internal echo pattern, using the movement's lift angle (the arc through which escapement and balance are engaged) as a calibration constant. Most modern Swiss levers use 52 degrees; many ETA calibres 50; co-axial and vintage designs differ. Enter the wrong lift angle and the amplitude reading — though not the rate — is systematically wrong, which can make a tired movement look healthy. When a seller shows you a timing result, the machine's settings are a fair thing to ask about.
Reading a timing result
A timing trace gives four figures. Daily rate in seconds (positive fast, negative slow). Beat error in milliseconds — the asymmetry between tick and tock, ideally under 0.3 ms, reflecting how perfectly the balance's impulse jewel centres between swings. Amplitude in degrees — how far the balance swings; 250 to 310 fully wound is healthy, while readings below about 200 point to degraded oil, a tired mainspring, or worse. And the beat rate confirms the calibre is what it claims. The skill is reading them together: good rate at low amplitude is a movement compensating, not a movement healthy; a clean flat trace in one position that scatters in another is an adjustment or wear story; beat error that changes with position suggests play somewhere it shouldn't be.
For buyers, requesting a multi-position timing printout on a significant pre-owned purchase is legitimate and informative. A movement that tests well flat but spreads 15+ seconds across positions has not been adjusted — not necessarily disqualifying, but worth knowing, and worth pricing. The half-hour of a watchmaker's bench time this costs is the cheapest insurance in collecting.
The goal is not to chase perfect numbers. It is to understand what the numbers say — about the movement's adjustment, its service history, and whether the watch will behave consistently in the positions it actually spends its life in. A timing machine cannot tell you whether a watch is good. It can tell you, precisely, whether it has been cared for by people who knew what they were doing.