Why we stopped scoring games out of ten
A ten-point scale packs how badly a game runs and how well it is written into one number. Readers cannot decide from that. Seven dimensions later, the complaints after purchase are more accurate.
Start with the conclusion: we did not get more generous, and we did not get stricter. We stopped pretending one number can summarise a game.
One number cannot hold three things
The trouble with a ten-point scale is that it reads too well. An 8 means "worth buying" and a 6 means "average", automatically. The mapping is so smooth that it covers every disagreement underneath.
Two games we both scored 7.
The first has above-average writing and presentation, but a rough PC build. The 1% low sits under 30 and mid-range cards stutter.
The second looks plain and tells a flat story, but it is technically solid. It holds 60 fps on anything, and it ran 40 hours without a crash.
Those two mean opposite things to a reader. One suits a good PC and an interest in story. The other suits modest hardware and a wish to relax. On the review page they stand side by side at 7, with nothing to separate them.
This is not a precision problem. Move to a hundred-point scale and the gap between 9.1 and 8.9 still means nothing. The dimensions are the problem, not the increments.
Seven dimensions instead
Every game page now records seven: performance, stability, amount of content, writing and presentation, controls, value, and post-launch support.
Each gets one of four levels rather than a number: excellent, good, fair, poor. Each needs a concrete line of evidence. Performance, good: the 1% low holds at 72 on 1080p high, but drops to 48 with ray tracing on.
Four levels rather than ten is deliberate. Ten invites arguments about 7.5 against 7.8. Four forces a call: is this good or fair? You have to pick a side.
Only performance is objective, in the sense that it can be measured. The other six carry judgment. We no longer hide that. Split the judgment into parts and show the evidence, and a reader can take only the parts that apply to them.
The score stays, demoted
A number still appears on the page. It is an internal index used mainly for ordering lists such as Top Rated. It is an index now, not a conclusion.
What a reader needs is the seven-row table. If all you care about is holding 60 fps, the performance row is enough and you can skip the other six. A ten-point scale cannot do that.
What changed
Three months in, two things moved.
Bounce rate on game pages fell. We are not publishing the figure, but the direction held: people stay longer, and more of them open the methodology page. They are checking the work rather than reading the verdict.
Post-purchase complaints changed shape. They used to read "you gave it an 8 and it stutters". Now they read "you marked performance fair, I bought it, it stutters, now I know". The first is a broken expectation. The second is an accurate one, and an accurate expectation is what a review is for.
What this costs
It takes more people. Seven rows plus evidence roughly doubles writing time, and you cannot bluff a row with a feeling. Every line needs something behind it.
It also travels worse. There is no "9.2 masterpiece" to screenshot and share, which is a real loss on social platforms.
We accept both costs. A score exists to help a reader decide. It is not a marketing device.
