Scores
The midpoint of the scale is not the middle of the results
Most tools centre their output somewhere between six and seven, and reading a six as below average is the most common misreading there is.
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
Five is the mathematical midpoint of a ten-point scale. It is almost never where a rating tool's results actually centre. Most tools cluster their average result somewhere between six and seven, which means a 6 is often close to average and a 5 can already sit below most of the population.
Two different fives
The scale midpoint is a property of the number line: half of ten is five, and nothing about that requires a tool's population to centre there. The distribution centre is a property of the tool's actual results: where most submissions actually land, once you look at a real sample of them. These two get conflated constantly because the same digit sits in both roles on the page, and only one of them tells you anything about your own result relative to everyone else's.
A reader who assumes the two are the same treats a 6 as "one below the middle" when the tool's real middle might be sitting at 6.4. That reading makes a roughly average result feel like a below-average one, for no reason connected to the actual submission.
Why tools shift the centre upward
A tool that routinely returns low numbers loses users faster than one that does not, and the mapping step between the model's internal judgement and the printed number is a place where that pressure can be absorbed without touching the assessment the model itself is running. Human rating systems drift the same way with no tuning at all: Filippas, Horton and Golden (2019) describe raters feeling pressure to leave "above average" ratings, which "in turn pushes the average higher". Generosity bias covers the mechanism and the incentive behind that shift in full; the point relevant here is just the consequence - it moves where the centre of the distribution sits, and it moves it up.
How to find where average actually sits, for a given tool
Three ways, in order of how much work they take.
Read the tool's own explanation, if it has one. Some tools state directly what an average result looks like or publish a rough distribution. This is rare and worth noticing when it happens - a tool willing to say "our average is 6.8" is telling you something most competitors will not.
Look at a sample of public results. If the tool shows other people's submissions and scores, even a rough eyeball of twenty or thirty gives a sense of where the bulk sits. Rate Cock publishes results with their full breakdown on public entries, which is the kind of sample you can actually read rather than guess at.
Compare your own repeated results to each other. Not a substitute for population data, but a run of your own submissions under standardised conditions at least tells you where your own number sits relative to itself over time, which is a different and still useful question from where you sit relative to everyone else.
The reading this avoids
A reader who does not know where a tool's centre actually sits has two ways to misjudge a result: treating an average score as disappointing because it is below the mathematical midpoint, or treating a merely above-centre score as exceptional because it is above 5. Both errors come from the same substitution - a scale property standing in for a population property - and both go away once the two are told apart.
There is a third, quieter version of this mistake: assuming the centre is fixed once you have located it. A tool's centre can move for the same reasons its whole calibration can drift - a retuned mapping, a shifted population, an edited rubric - so a location found this year is not guaranteed to hold next year, and the same three checks above are worth rerunning rather than trusted indefinitely.
Whose population the centre is even calculated against is its own separate question, one that changes what "average" is claiming to describe in the first place; the reference population behind a score picks that thread up. None of this is about size or measurement in any physical sense - that's a different property entirely, owned by an actual tape and a method, and where a scale's centre sits has nothing to do with what that kind of instrument would report. A rating's centre is a statement about a distribution of numbers, not a statement about bodies, and it is a different claim again from what a person reviewing submissions directly considers typical based on their own experience rather than a logged population.