Scores
Ninety-nine steps, same amount of information
Stretching a scale adds digits, not resolution; the underlying judgement did not get sharper because the display got longer.
Guides on Scores: How much weight a rating tool result deserves, Every part of a rating result, and what each one is for, What can and cannot be compared, and how
A tool that scores you 74 out of 100 looks like it did more work than one that scores you 7.4 out of ten. It did not. Both numbers came from the same underlying process, and stretching the display onto a longer scale changes nothing about what that process was actually capable of resolving.
Same claim, more digits
Multiply a ten-point scale by ten and every score gets longer without getting any more accurate. 74 and 7.4 are the same claim wearing a different number of zeroes. The internal value that produced either display was a single continuous estimate, rounded once for presentation - the choice of how many positions to round it to is cosmetic, made after the estimate already existed, and it cannot add information that was not already there.
A hundred-point scale can create an illusion of resolution it does not have. If two submissions land at 74 and 76, the gap looks specific and checkable - two whole points, out of a hundred, surely meaningful. Convert both back to the ten-point original they came from and the gap is two tenths, well inside the range two identical submissions can drift on repeat. The hundred-point display did not shrink that noise. It just gave the noise more room to look like signal. What actually narrows uncertainty is more evidence: in the NIST/SEMATECH e-Handbook of Statistical Methods, the interval around a mean gets narrower "as N increases," through repeated observations, and nothing in that formula responds to how many digits a display prints.
Why tools do this anyway
A longer number reads as more analytical, and "94/100" carries a certain authority that "9.4/10" does not quite manage, even though they are the same result. It is a presentation decision, not a methodological one - nothing about scoring on a hundred-point scale requires the underlying process to actually distinguish a hundred different outcomes, and almost none of them do.
What to do with a hundred-point score
Divide by ten and read it as you would any other rating: as a value with real resolution somewhere around a handful of distinguishable levels, not ninety-nine. The same caution about the decimal applies here, just multiplied - if a single decimal on a ten-point scale is already close to the edge of what the tool can reliably repeat, a hundred-point scale has not solved that problem, it has just hidden it behind more digits. Whether a tool prints a total on ten points, a hundred, or five stars, the honest question is what shape the number takes and what each shape can support, and a hundred-point scale is not automatically the more honest of the three - often the reverse.
None of the three satellites in this category solve the problem by using a different scale. What a model estimates in the first place sets the real resolution long before any display scale is chosen; an actual measured length carries its own precision independent of any of this; and a human reviewer rarely bothers with more than a whole number regardless. Rate Cock keeps its total on the conventional ten points rather than stretching it to a hundred, which at least avoids manufacturing precision the underlying six-axis process was never built to deliver.