Data methodology

What does the database hold?

Published tractor models

2,500

Makes

71

Laboratory test reports

177

Models with PTO power measured

138

Power figures

4,682

Specification values

116,280

Test measurements

4,252

Cited documents

3,204

Production years covered: 1913 to 2016.

Where does each figure come from?

Every stored value carries a source record: the document, how the value was extracted and when it was retrieved. A value without a source is not stored. Numbers are kept as a number, a unit and the precision the source printed, and converted for display; the figure as printed is kept beside it. Beside each figure a page names the source and what kind of document it is; the page’s Sources block cites each document in full, with its licence.

Why are measured and claimed power shown separately?

Gross engine power, net engine power, PTO power and drawbar power are different quantities, and a figure measured by a testing laboratory is a different kind of figure from a manufacturer’s claim. Each is stored with its quantity, basis (measured or claimed) and the test standard where the source states one. They are never merged into one horsepower number. Where a laboratory figure and a claim differ, both are shown.

What happens when two sources disagree?

Where two sources state the same quantity for the same machine, the figures are compared like with like: the same quantity, the same basis and the same test condition. A manufacturer’s claim and a laboratory measurement that differ are not a disagreement; the gap between them is the point of showing both. Two measurements at different test conditions are not a disagreement either.

Where two sources genuinely disagree, we do not pick one quietly. Both figures stay on record, the value is marked as disputed on the page, and each source is named beside its figure. Disputed values on published pages now: 326, on 55 models.

Torque figures and fluid capacities are held to a higher standard, because a wrong torque figure can damage a machine. They are shown only when two independent sources agree (different publishers, neither copied from the other) or an editor has checked the figure against its source, shown as “Verified by WhichTractor editorial”.

Which models are published?

A model is published only when every item of the buyer core set that applies to its era is on record, each with a source:

• Production years
• Engine maker
• Cylinders
• Displacement
• Power (any standard, as stated)
• Drive
• Transmission (gear counts or type)
• Mass
• Rear PTO speeds
• Hydraulic lift capacity or pump flow (models from 1950)
• Wheelbase or overall length

Records that fall short are not published at all: no page, no listing, no sitemap entry.

How is a WhichTractor size assigned?

A size (compact, utility, standard-size, high-power) is our classification, not a manufacturer’s category. It is assigned by a written, versioned rule from the machine’s power relative to other tractors of the same era band (pre-1940, 1940-1959, 1960-1979, 1980-1999, 2000-2015, 2016-), so forty horsepower in 1950 and forty horsepower today are not given the same size. A machine within two percentile points of a boundary is marked as sitting on it.

How do we decide what a machine is?

Every page on this site sits under /tractors/, and for most of the corpus that was an assumption rather than a finding: the record said nothing about what the machine was, and nothing said it was not a tractor, so it published as one. An assumption that decides what gets published should be visible, so each page now says which of three things it is.

• The source says so. The reference work states it in a sentence about that machine: of the Güldner G 40 W, that it is a Güldner tractor of the 1960s. The sentence is kept with the value in its own language, and is not reproduced here.
• We derived it. The machine carries a sourced three-point linkage — a hitch category, a lift capacity or a hitch control. A three-point linkage is the functional signature of a tractor. Every input is itself sourced and the rule is the same for every machine.
• Unspecified. We looked and cannot say. The page says so rather than defaulting.

We do not use the power take-off for this. A PTO is not enough on its own: transporters and motor mowers have one, and an earlier version of this rule called seventeen of them tractors before it was tested against machines we already knew were not.

How is a machine classified?

A classification (drive, configuration such as orchard or crawler, transmission type, PTO type, operator station) is recorded only where a source has a sentence that says what this machine is. A word in a list of optional equipment is not enough, nor is a comparison with other models, a price or market position, a description of a different machine, or an inference from dates. Where no such sentence exists, the value is “unspecified”, and the page says so rather than guess. The same holds for abbreviations and model suffixes: they are never expanded without a source.

How are two tractors compared?

A comparison table sets the facts of two published models side by side, and every figure shows its basis (measured, claimed or declared), so a claim is never read against a measurement. A written verdict is a separate page and is given only when the two machines are real alternatives: both have a stated, comparable configuration, or their headline power is within 20%, their unladen mass within 25%, they share a drive and their production years overlap.

A verdict never names a winner on a difference smaller than the figures can support. Two laboratory PTO figures measured at the same test condition must differ by more than 5%; two manufacturers’ claims by more than 10%, because claims come from different makers, eras and unstated standards and are often converted and rounded. Below those gaps the verdict says there is no meaningful difference.

Where do serial number facts come from?

Serial number ranges, first and last machines, and parts or specifications that changed at a serial number are recorded exactly as the source prints them, prefixes and spacing included. A number is called a tractor or engine number only where the source’s own word says which; otherwise it is shown as a serial number without that claim. Lines that are ambiguous, list alternatives or describe two changes at once are held back rather than interpreted. Published models with serial number facts: 74 (269 facts, 43 recorded changes).

What happens when two records may be the same tractor?

Two records are merged only when a written rule finds the same designation, overlapping years and no conflicting facts; otherwise the pair waits for a person. While it waits, both published records say so and link each other. Pairs waiting now: 281.

What sources are never used?

Other specification websites and compiled databases are never used as a source of figures, only, at most, as a list of machines to look for. Where a site’s terms forbid automated collection, where the site blocks us, or where its owner asks us to stop, we stop. Prose, photographs and brochure images are not copied; photographs on the site carry their own licence and credit. We never generate an image of a specific tractor.

How should this data be cited?

Cite the page the figure appears on, and where you can, the original source the page names beside it: a laboratory report number or a publication is the stronger citation. For the database as a whole:

WhichTractor tractor specification database. Data methodology. whichtractor.com/data-methodology/, data as of 2026-09-15.

Text drawn from the Traktorenlexikon on Wikibooks is shared under CC BY-SA 4.0 and must keep that attribution. Found an error? Every correction needs a source too: how corrections work.