Scores
Generosity is not quality
People prefer the tool that rated them highest, which is exactly the tool with the least useful scale.
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
No - a tool that scores you higher is usually just more generous, and a generous scale carries less information, not more accuracy. People tend to trust whichever tool rated them best, a natural instinct and a bad selection criterion, because what produces a flattering number and what produces a useful one pull in opposite directions.
Why the preference is backwards
A tool scores high for one of two reasons: because the submission genuinely earns it, or because the tool's scale is set generously across the board. From the outside, on one result, you cannot tell which. The only way the generous scale ever gets exposed is by comparison with other results - your own over time, or other people's - and a single flattering number gives you neither.
Meanwhile, tools under commercial pressure have every incentive to let their scale creep upward and none to let it creep down, because a tool that scores low loses the user who just paid for the result. Preferring the highest-scoring tool selects for whichever one has drifted furthest in that direction, which is the tool whose number carries the least information, not the most.
What a generous scale costs you
A scale where most results land in the top two points cannot distinguish much within that range - that compression is the ceiling effect, and it is a direct cost of generosity, not a separate issue. Measurement research treats this as a testable defect: the Terwee et al. (2007) criterion, as applied by Wamper and colleagues (2010, Acta Orthopaedica), counts 15% of respondents at the top score as the maximum acceptable, because past that "a person may continue to improve, but the test does not capture that improvement." If everyone scores 8 or 9, an 8.4 and an 8.9 are not meaningfully different results; the scale ran out of room to say so. A tool you picked because it made you feel good has, by the same mechanism, made itself worse at telling you anything.
What to look for instead
Spread and consistency are the properties worth selecting on. Spread: does the tool's distribution actually use the bottom half of its scale, or does everyone land in a narrow band near the top? Consistency: does the identical submission, resubmitted, come back close to the same number, or does it wander? Neither of these can be judged from one result on one tool - they require either public distribution data or your own repeated submissions, which is more effort than reading a single flattering number, and also the only way to tell a generous tool from an accurate one.
This is a narrower point than the mechanics of why generosity happens - that mechanism is its own subject - and a different failure from a tool overstating certainty about a rank it never earned, which is more the territory of a human reviewer's confidence than an automated scale's. It has nothing to do with measurement either - a tape does not have a generosity bias to correct for, only a rating scale does. The underlying model rarely changes between a generous tool and a strict one; what usually differs is the layer built on top of similar shared components, which is exactly where the scale gets set. Rate Cock publishes enough spread across its public entries that the distribution is checkable rather than assumed, which is the actual test this piece is arguing for.