Written out fully so that a reader can repeat any test in this publication, get a different result, and hold us to it. The procedure has been amended three times, each time because a reader demonstrated something we had missed.
Nothing is published before a full in-game year has been observed, or forty hours of play where the game has no calendar. Every figure comes from a paper log kept during play rather than from instrumentation.Every system is placed in one of three categories, and the category is printed beside the score on every page. Simulated means the weather produces consequences that persist beyond the condition itself and apply to the world as well as to the player. Partial means consequences exist but run in one direction only, most commonly as a benefit with no matching cost. Cosmetic means the weather changes the image and the sound and nothing measurable besides.
There is deliberately nothing between the categories. A system either changes what happens or it does not, and the reviews carry the nuance in prose rather than in a fourth label.
The core of the procedure. An action is performed in clear conditions and then repeated in each available weather state, along the same route, from the same position, with everything else held as constant as the game allows. The difference is recorded, and any action showing no difference is repeated at least ten times before being logged as unaffected.
Where a game exposes a number — a detection distance, a movement speed, a growth timer — we prefer to derive it by repeated observation rather than by reading a file, because the value in memory and the value the player experiences are not always the same thing.
Whether the weather applies to anybody other than you. Fog that blinds the player and not the guards, rain that soaks the player and not the livestock, wind that pushes your boat and nobody else's: these are the most common failures in the file and the least often noticed, because the visual effect is so persuasive that a player assumes the system behind it is complete.
Testing this requires patience and no tools. Approach the same non-player character in four conditions and record when they react. Follow another vessel through a deterioration and see whether it behaves as though the weather exists. Leave animals exposed and observe them.
Measured with a stopwatch across at least twenty complete transitions, from the first visible precursor of a change to the point at which the new condition is fully established. We record both the total cycle length and whether transitions are gradual or snap between states, since a system that changes instantly cannot be anticipated regardless of how long each condition lasts.
Cycle length is the second most predictive figure we take. Nothing running faster than about fifteen minutes has yet proved forecastable, and forecasting is the most rewarding thing a sky can offer.
We note every environmental indicator a player can learn to read — cloud base, swell behaviour, the movement of loose snow, animal behaviour, the pitch of wind — and we record how long it took us to learn to use each one. A system that can only be read from a panel is a weaker system than one that can be read from the horizon, even when both contain the same information.
A review score for the entire game. The weather classification is a separate judgement printed beside it and never folded into it, which is why a game with a purely cosmetic sky can score well and a game with an exemplary one can be marked down for everything else. Two entries in the current file are exactly those cases.
All observations are made on an unmodified installation at the game's own default settings, since a weather system is what shipped rather than what a community rebuilt. Where a mod materially improves a system, an article may say so in prose; no logged figure is ever taken from a modified installation.
Weather systems are patched often, and the most consequential changes are rarely mentioned in a changelog. When a patch alters something we have logged, the test is repeated and the review is annotated with the date and both results, leaving the original visible, because the behaviour at release was real and readers who bought it then experienced it.
One tester, one machine, one set of habits, and a strong bias toward games whose weather looked interesting enough to examine. Aggregate figures for the file as a whole should be read with that selection in mind. Consequence tests can only find effects we thought to look for, which is why the procedure has been amended three times by readers rather than by us. And a full in-game year is enough to see a system's rules but not always enough to see its rarest state.
Every game logged here was bought at retail. No developer, publisher or platform has supplied a copy, funded an article or seen anything before publication. There is no advertising, no sponsorship, no affiliate arrangement and no storefront link anywhere on this site.
Figures are corrected in place with a dated note whenever a reader demonstrates an error or a patch changes behaviour. There is no address, form or submission route on this site by design; corrections arrive because readers repeat these tests publicly and describe their method, which is a higher standard of evidence than a message would be.
