Romaji to Kana Converter: React & Next.js — Chapter 4, Exercise 3
=====================================================================

TASK
Explain, in your own words, why the measured ratio (38,000x vs. the
Astro course's own ~10,353x) is a less meaningful comparison between
the two variants than the two raw absolute numbers are.

SOLUTION

A ratio is only as stable as the two numbers it's built from, and
one of those two numbers here -- the direct-call time -- is
extremely small to begin with, on the order of a fraction of a
microsecond. Timing measurements at that scale carry real,
unavoidable noise: exactly how fast a tight loop of function calls
runs in a real run depends on small, genuinely random factors --
what else the CPU happens to be doing that moment, how the JIT
compiler has warmed up, minor scheduling jitter -- that shift the
result by amounts that are tiny in absolute terms but large as a
fraction of an already-tiny number. Dividing a large, comparatively
stable number (the real HTTP round-trip cost) by a small, noisy one
(the direct-call cost) means all of that small-number noise gets
carried straight into the ratio, and even gets amplified, since a
proportionally small change in a sub-microsecond baseline can swing
the ratio by thousands.

The two raw absolute numbers don't have this problem nearly as
badly. The real HTTP round-trip cost -- roughly 14,000-15,000
microseconds in both this course's own measurements and the Astro
course's own separate measurement -- is dominated by a real,
comparatively large and stable cost: the actual network/socket round
trip. That number stays close to the same figure across multiple
runs, multiple words, and even across two completely different
server implementations written for two different courses. The
direct-call number, while individually noisier in relative terms, is
still unambiguously "a small fraction of a single millisecond" no
matter which specific sub-microsecond value comes out of a given run.

So the real, meaningful comparison between the Astro course and this
one isn't "is 38,000x bigger than 10,353x" -- that's really just
measuring how much noise happened to be in each course's own
sub-microsecond baseline that day. The meaningful comparison is that
both courses independently measured the real HTTP round-trip cost at
roughly the same order of magnitude (right around 14 milliseconds),
against a direct call that's negligibly fast either way -- and that
shared, stable finding is what actually supports the shared verdict
(client-side wins), not the specific multiplier either course happens
to report.

WHY THIS WORKS AS AN ANSWER
----------------------------
It explains the real statistical reason a ratio built from a large,
stable number and a small, noisy one inherits and amplifies that
noise, and correctly identifies that the two courses' own real
agreement on the large, stable HTTP-cost figure -- not the ratio --
is what actually makes their shared verdict trustworthy.