Best For Whom
Every "best tool" argument is secretly running a small regression. Score equals how good the tool is, plus how well it fits what you already have, plus noise. Most comparisons spend all their effort isolating the first term and none on the second, as if fit were measurement error to be averaged away rather than a real variable with its own weight.
Take the choice between a self-hosted sync tool and a paid one. The self-hosted option has no server, which sounds like a feature until you notice what it actually means: nothing moves your files unless two machines happen to be awake at the same time. If you already own something that's always on, a Pi humming in a closet, a cheap box you were going to rent anyway, that constraint costs you nothing. The sync tool is free, it's yours, and the paid competitor is now just a worse version of a thing you already have running.
If you don't own that always-on machine, the equation flips completely. Now "free" sync software means you first have to acquire, configure, and keep alive a piece of infrastructure you didn't need before, just to unlock the thing the paid service does by default. The self-hosted tool didn't get worse. Your denominator changed. Two people can read the identical feature comparison, apply identical reasoning, and correctly land on opposite tools, because the thing that decided the outcome was never in the comparison at all.
Distro choice works the same way. Arguments for one package manager over another tend to be argued as if there's a single ranked list, this one resolves dependencies better, that one boots faster, and the list is treated as portable, something you could hand to a stranger and expect the same verdict. But half the actual cost of a distro is what you already know how to fix at 2am, what your existing config already assumes, what muscle memory your hands have for recovering from a broken update. A tool that's objectively cleaner but demands you relearn your entire recovery instinct can be the wrong choice even when every review says it's the right one, because the review's scoring never had a term for "you, specifically, already know how to fight the other one."
None of this is an argument for relativism, everything's a wash, whatever gets you through the night. Fit to your setup is not a mood. It's countable: the marginal cost of the hardware you'd need to buy, the hours of recovered muscle memory you'd throw away, the maintenance capacity you actually have this year versus the capacity a tutorial assumes you have. Those numbers are as real as a benchmark score, just personal to the chooser instead of global to the tool. Treating them as noise doesn't make a comparison more objective, it just quietly substitutes the reviewer's setup for yours and calls the result universal.
The reason best-tool-wins culture keeps doing this anyway is that a leaderboard is easier to write than a function. "X beats Y" fits in a headline. "X beats Y if you already have Z, otherwise Y beats X" does not, and it also doesn't generate the kind of confident argument that's fun to have online. A single ranked list gives everyone something to agree or disagree with. A conditional one just describes two people who are both right.
So when two people land on different tools after doing the same homework, the interesting question isn't which one made the mistake. It's what term in their equation you're missing.