Methodology

How we score, and where the numbers come from

Frame rates here are not measured by us, and we label how each one was arrived at. A figure marked <em>measured</em> comes from a published run on that exact machine. A figure marked <em>representative</em> is derived from the GPU and its power limit (TGP) and cross-referenced against published reviews - a sound estimate of what that configuration delivers, but not a measurement of that specific laptop. Unless stated otherwise, numbers are 1920x1080, High preset, upscaling off. Where a laptop ships in several power configurations we use the wattage of the exact SKU listed, because a 140W RTX 5070 and a 115W one are not the same product.

The score is computed, not awarded

Nobody here decides that a laptop "feels like an 8". The score is calculated from four measurable things and the breakdown is printed on every review, so you can see exactly which part carried it - and disagree with the weighting:

Performance is the mean frame rate across every game we have figures for, on a curve where 60fps scores about 4.9 and 200fps scores 10 - because the difference between 40 and 60fps matters far more to a player than the difference between 180 and 200. Value is frames per £100. Screen counts pixels above 1080p, refresh rate and panel type. Portability is weight.

The anchors are absolute rather than relative to whatever we happen to have reviewed, so a 7.9 means the same thing next year as it does today, and adding a laptop to the database doesn't silently re-rank everything else.

The same games, every time

Only these 5 games count towards a score: Cyberpunk 2077, Baldur's Gate 3, GTA V, X-Plane 11, Dota 2. Different reviewers test different suites, and a mean over one set of games isn't comparable to a mean over another - so if we let each laptop be judged on whatever it happened to be tested with, the leaderboard would be measuring the reviewer as much as the laptop. Figures for other games still get published on the review page. They just don't move the number.

We didn't pick that list by taste. Test suites get rotated - the newest, most demanding titles only appear on the newest reviews - so a suite built from the games we'd like to rank on turned out to be held complete by exactly one machine in our database, which would have made the leaderboard a comparison of publication dates. So we computed it instead: for every possible set of a given size, how many laptops have all of it. This five-game set is the widest overlap available, and it's the one the scores are built on. As coverage of newer titles grows, the suite will move, and when it does we'll say so here.

When we don't give a score

Two things are required. A laptop missing either still gets a full review and every figure we hold - it just doesn't get a number, and it can't be a budget pick or appear in a head-to-head. We'd rather show you a gap than fill it with something we made up.

1. The complete benchmark suite. Sourced frame rates for all 5 of those games, not most of them. A partial suite defeats the entire point: averaging one machine's Cyberpunk 2077 and Baldur's Gate 3 against another's Cyberpunk 2077, GTA V and X-Plane 11 compares two test suites, not two laptops. Reviews of machines short of the set name the exact games we're missing.

2. A confirmed UK price. Value for money is 30% of the score, and you can't compute it without a price. The tempting alternative is to score the components we do have and rescale - but that quietly rewards the machines we know least about, because value is the only component that can drag a score down. We know because we tried it: under that rule every laptop above 6.0 was one we couldn't price, and every laptop you could actually walk out and buy ranked below all of them. So a machine we can't find on UK sale in the configuration we tested goes unscored. Its frame rates still rank on the leaderboard, which needs no price - a score is a buying recommendation, and there isn't one to make about something that isn't for sale.

Measured, or representative?

We label every frame rate as one of two things, because the difference matters and most sites blur it:

• Measured - a published benchmark run on that exact machine, credited to whoever ran it.
• Representative - derived from the GPU and its power limit, then cross-referenced against published reviews. This is a sound estimate of what that configuration delivers. It is not a measurement of that specific laptop, and we won't present it as one.

Right now the figures carried over from our launch dataset are representative, and every page that shows them says so. As measured runs replace them, the labels change with them.

What we don't do

We don't run benchmarks ourselves. We think aggregating and crediting published work is more useful than adding one more set of unverifiable numbers - but it does mean you should treat every figure as indicative of the configuration named, not a guarantee for the exact unit you buy.

Who writes this