Every page on this site is organized around a reference number. One number, one file, one watch. That is the promise a reference makes, and this dossier is about the places where the promise quietly breaks. Because a reference number is a container, not a timestamp. The factory assigns it when a design is approved and retires it when the design is replaced, and everything that happens in between, every dial revision, every lume change, every caseback rewording, happens inside the number without disturbing it. Some containers hold a year or two of production and the name fits like a caseback. Others hold twenty years, and inside those twenty years are watches so different from each other that collectors had to invent an entire unofficial science to tell them apart.

The twenty-year number

Exhibit A is the GMT-Master 1675, one of the longest-running references Rolex ever issued. It entered the catalog in 1959, fixing the exposed crown and fragile bezel of the 6542 by adding crown guards and growing the case to 40mm, and it did not leave until 1980, when the 16750 finally replaced it. Twenty-one years, one number. Now count what changed inside it. The crown guards started as the sharp pointed horns collectors call cornino and were later rounded off, and the pointed era alone contains documented sub-generations. The dial began as glossy black with gilt printing and a chapter ring, lost the ring in the mid 1960s, then went matte with white printing for the rest of the run. The lume switched from radium to tritium in 1963 as radiation rules tightened. The movement changed from caliber 1565 to 1575 around 1964 to 1965. One researched count puts the tally at eight distinct lacquered dial types and nine matte ones, all wearing the same four digits.

Rolex renamed none of it. The factory's paperwork says 1675 from the first watch to the last, so the hobby built its own paperwork in the gap. Matte 1675 dials are sorted today into collector designations that run from Mark 0 up through Mark 6, the crown guards have their own Mark sequence, and none of this nomenclature appears in any Rolex document. It is unofficial stratigraphy, scholarship layered onto a number that stopped carrying information, and like all scholarship it gets argued about and revised, which is exactly the condition this register's founding dossier, Watches That Disagree With Their Makers, was built to record. The GMT-Master family file can tell you where the 1675 sits in the line. It cannot tell you where your 1675 sits inside the 1675. For that you need the strata.

The maker who tried to fix it

Omega saw this problem and did something almost no maker does: it dated its own sub-variants. The Speedmaster 145.022, the first Moonwatch reference built around the caliber 861 after the 145.012 ended the 321 era, carried a suffix stamped inside the caseback, and the suffix was a year. 145.022-68. Then -69, -71, -74, -76, -78. The house itself was admitting that one number would hold a decade and offering a decoder for it. The -68 is the transitional watch, still wearing the applied metal logo of the 321 era over the new movement. The -69 spans the moon landing. The -71 brought the medallion caseback with its Flight-Qualified engraving. The -74 is where the stepped dial gives way to the flat one, and the -78 is the last suffix before Omega abandoned the experiment entirely and let the number run bare.

And here is the register's favourite part: even the maker's own dating system turned out to be a container. The suffix marks when the case reference was issued, not when the watch was assembled, so a watch with -69 stamped in its caseback can carry an archive extract reading June 1970, and watches stamped -74 were still being cased in 1976. The bezel font changed from the dot-over-ninety to the dot-next-to-ninety partway through the -69 run, at a serial number, not at a suffix. The scholarship has had to revise itself in public: for years the -71 was recorded as the last stepped dial, until enough honest -74s surfaced with steps that the record moved. And the number kept mutating after the suffixes died, becoming 145.0022 in the early 1980s, then living inside the 3570.50 and today's 311.30.42 product codes, four numbering regimes wrapped around what the genealogy dossier shows is one unbroken watch. Omega built the most honest reference system in this register and it still could not make the number tell the time.

Two printed words

Seiko's version of this story is the best documented accident in the file. The 6139-6000, one of the finish-line watches from the 1969 automatic chronograph race, spent its first years with Water 70m Proof printed on the dial and Proof on the caseback. Japanese and American trade rules then forced the language to change, so Proof became Resist on dials around mid 1970, casebacks moved from Waterproof through Water Resist to Water Resistant, and by about 1973 the water rating left the dial altogether. Nobody at Seiko was building a dating system. They were complying with labeling law. But the result is that two printed words now date a 6139 to within a couple of years, and the transition is messy in exactly the way real strata are: the American market sub-variant had flipped to Resist by March 1970 while rest-of-world dials were still leaving the factory marked Proof that August, and transitional watches exist with a Proof dial inside a Resist caseback.

The strangest consequence sits on the most famous dial in the family. The yellow chronograph Colonel William Pogue wore into orbit is pinned to its production window by its own text: his watch read Water 70m Resist, which places it in the 1970 to 1972 band, which tells collectors which sub-variant he actually wore, and the hobby's definition of a True Pogue rests on those two words. The 6139-6002 file carries the same sediment. One reference, and the dial printing is doing the work the number never could.

Not a vintage problem

If this sounds like a disease of the 1960s, the modern file says otherwise. The GMT-Master II 16710 ran from 1989 to 2007 and holds three bezel colours, two calibers and three lume chemistries under one number. The dial text at six o'clock reads Swiss T under 25 for the tritium years, then changes as Luminova arrives, and the sources this register checked split on whether that happened in 1998 or 1999, a one-year disagreement the file records rather than resolves, because recording splits is the house doctrine. The last watches carry caliber 3186 in place of the 3185. None of that is visible in the number. A first-year 16710 and a final-year 16710 are separated by eighteen years, a movement, and a lume chemistry, and their casebacks agree with each other completely. Then the 116710LN arrived and the container was sealed.

How to read inside the number

So the doctrine, for any reference whose file spans more than a few years. The reference tells you what the watch is. The serial tells you when it was born, which is why the Rolex serial lookup, the Omega serial lookup and the Seiko serial decoder exist on this site, and why they are the first stop, not the last. Then the strata confirm or contradict the serial: dial text and printing, bezel fonts, caseback wording, lume material, movement caliber, case details down to the shape of a crown guard. When serial and strata agree, you have a watch. When they disagree, you have a question, and sometimes a watch assembled from periods that never met, which is its own entry in the family-splitting story of how these lines are told. The unofficial Mark systems are the hobby's answer to a gap the makers left, which means they are living documents, revised whenever a documented example breaks the pattern, and this register treats them the way it treats every contested record: useful, citable, and never final.

The number on the caseback tells you what the watch is. It has never told you when.