Blog
How we compute the Linux Readiness Score
The Linux Readiness Score is a number we compute from a published formula, not a number we feel. A hand-set “we think it’s a 7” is not citable, because nobody can argue with taste. A formula recomputed from per-distro hardware data is citable, because anyone who disagrees has to argue with the arithmetic. This post is the short version of how it works and why it is built the way it is. The full specification, with every weight and its justification, is on the methodology page.
What goes in
The score is built from two things already recorded for each laptop, per distro:
- The grade. The human verdict on a clean install: does it just work (
out-of-box), need one or two documented fixes (minor-tweaks), or have significant gaps (problematic). - Six hardware components. Wi-Fi, Bluetooth, fingerprint, webcam, audio and suspend, each marked
ok,tweak(works after a known fix),broken, orunknown.
Both are computed per distro, combined into a per-distro sub-score, and the model’s published score is the best supported distro it has. Best, not average, on purpose: a machine that runs perfectly on Fedora but is untested on Arch is a good Linux machine, and averaging in the untested distro would punish a laptop for the breadth of the test matrix rather than describe the hardware.
How the pieces combine
The grade maps to points: out-of-box is 10, minor-tweaks is 7, problematic is 3. The gap between out-of-box and minor-tweaks is deliberately 3 points, the cost of having to know a fix exists at all.
The six components are not equal. A laptop you cannot suspend is a laptop you cannot close the lid on; a missing fingerprint reader is just a missing convenience. So suspend carries the heaviest weight (0.28), Wi-Fi is close behind (0.26), then audio (0.16), webcam (0.14), Bluetooth (0.10), and fingerprint last (0.06). Those six weights are the load-bearing opinion of the whole metric, which is exactly why they are published and open to public argument rather than buried.
A per-distro sub-score is 45 percent grade and 55 percent hardware. Hardware gets the larger share because it is the falsifiable evidence; the grade keeps a heavy 45 percent because it catches things the six components do not encode, like GPU-driver pain, thermal throttling on battery, or “it boots but the trackpad is unusable”. Neither term is allowed to fully override the other. The Dell XPS 13 9350 is the textbook case: only the webcam is broken, everything else is fine, but the grade is problematic, and the 45 percent grade anchor is what stops it floating up to a 9.
Why “unknown” never helps the score
This is the rule that matters most for honesty. A component nobody has verified is excluded from the hardware term and counted against confidence. It is never scored as working. A machine nobody has tested is not a good Linux machine; it is an untested one, and the number has to say so. If a model has no distro that is both graded and has at least three of six components known, it does not get a number at all. It renders as “Not rated”, not as a zero and not as an inflated guess.
There is one honest subtlety. On the EU Linux vendors, several machines ship with no fingerprint reader at all, by design, for example the Tuxedo InfinityBook Pro 14 Gen10 and the Star Labs StarBook Mk VII. “No reader exists” and “we did not test the reader” are different facts that the schema currently serialises the same way. The conservative default until that is disambiguated in the data is to treat a fingerprint unknown like any other unknown: excluded from the term and counted against confidence. When in doubt the score degrades confidence, it never inflates. The full reasoning, including the proposed schema fix, is in the methodology page.
Why it is built this way
Three constraints drove every decision. It has to be deterministic: same data in, same number out, forever, no model-of-the-week. It has to be honest about ignorance: unknown must never raise the score. And every weight has to be defensible in one line a stranger can attack in public. A score that fails any of those is a marketing badge, not a measurement, and a badge is not something anyone cites back to you.
That last point is the entire reason the formula is public. The value of the score is not the digit; it is that the method behind the digit is open, dated, and falsifiable. If the per-model data goes stale the score is wrong, and a wrong dated number that shows its working is still more useful than a confident undated one. For the complete formula, the per-distro roll-up rules, and the justification behind every weight, read the methodology page.
FAQ
What is the Linux Readiness Score? A 0 to 10 number computed from a published formula, blending the per-distro install grade with six weighted hardware components. It replaces a hand-set editorial number with a deterministic one, so a disagreement is an argument with the arithmetic, not with taste.
What goes into the score?
Two recorded inputs per distro: the install grade (out-of-box, minor-tweaks, problematic) and six components (Wi-Fi, Bluetooth, fingerprint, webcam, audio, suspend). They combine 45 percent grade, 55 percent hardware, and the model takes its best supported distro.
Why doesn’t an untested laptop get a high score? Because unknown never helps. Untested components are excluded from the hardware term and count against confidence; a model with no sufficiently-tested distro shows “Not rated”, not a zero or a guess. An untested machine is not a good Linux machine, just an unknown one.
Why is suspend weighted so heavily? At 0.28 it is the heaviest component because broken suspend means the battery dies in a bag or the machine never wakes, it is the top real-world Linux-laptop complaint, and it is the hardest fault for a user to work around. Wi-Fi is close behind at 0.26.
Where is the full methodology? On the methodology page. It lists every weight, the grade-versus-hardware split, the per-distro roll-up rules, the “best supported distro” choice, and the fingerprint “absent by design” problem with its proposed schema fix.