The method, in full
Two players, two drafts, one named disagreement
Everything on this site follows the same sequence, and the sequence has not changed since early 2022. It is not clever and it is not fast. What it does is make our blind spots visible: when two people who work in the same room reach different conclusions about the same build, the difference tells you something a single write-up never could.
The rest of this page sets out each step, the hardware we play on, what we do about corrections, and the handful of things we deliberately refuse to do.
Sequence
From queue to publication
Seven steps. Steps three and five are the ones that most often send a game back to the queue.
-
Into the queue
A game enters the queue either because one of us wants to play it or because a reader made a specific argument for it. Publisher approaches go in the same queue with no priority. The queue is ordered by who has time, not by release date.
-
Pairing
Two of the four of us commit to it. We deliberately avoid always pairing the same two people, because settled pairs develop shared assumptions and the disagreements get quieter over time. Where possible one player is unfamiliar with the genre.
-
Version lock
Both players note the build number before starting and agree not to update mid-run unless a patch is unavoidable. If one player ends up on a materially different version, the pair is void and we start again once both can match.
-
Separate play, overlapping fortnight
Unless the game is co-operative by design, both runs are solo and private. Both must finish within fourteen days of each other. No discussion of the game during this window beyond “still going” or “stopped early”.
-
Independent drafts
Each player writes their column without seeing the other. Drafts go into a folder neither can open until both are filed. This rule exists because we used to talk about games while drafting and the columns quietly converged into a single house opinion.
-
The disagreement session
Both columns are read aloud in the same room. If a genuine disagreement exists, we write it down in one paragraph pointing at a specific mechanic, act or decision. If we cannot make it that specific, the verdict is recorded as a stalemate rather than a split.
-
Publication and hours logged
Both columns publish unedited except for clarity and length. Hours played, completion state and hardware go at the foot of every entry, because a fourteen-hour opinion and a sixty-hour opinion are different objects.
Verdict labels
Three outcomes, no numbers
Every entry closes with one of these. They describe the state of the desk, not the quality of the game.
Two players independently reached the same conclusion about who the game suits and why. This is the closest thing to a recommendation we produce, and it is still not a score.
Roughly a third of our write-ups land here. Consensus is less common than you might expect, and that is the whole reason the site exists.
A split means we agree about what the game does and disagree about whether it is a good idea. The disagreement paragraph names the exact mechanic involved, so you can decide which side describes you.
A stalemate means we cannot even agree on what the game is, usually because we played it in incompatible ways. Both accounts stand; neither is a recommendation.
| Period | Pairs published | Most common label | Note from the desk |
|---|---|---|---|
| 2022 | 14 | Consensus | Too many. We were pairing the same two people repeatedly and the columns had started to rhyme. |
| 2023 | 21 | Split | First year of rotating pairs and locking drafts. The change in distribution was immediate and obvious. |
| 2024 | 19 | Split | Introduced the rule that a disagreement must name a mechanic. Four entries were downgraded to stalemate. |
| 2025 | 23 | Split | Steady. Added hardware disclosure to the foot of every entry after a reader pointed out a performance claim we could not support. |
| 2026 so far | 11 | Split | Two stalemates already, both from management sims where our play patterns diverged completely. |
Hardware
What we actually play on
Nothing here is a review rig. Two of the four machines are inherited from previous jobs.
| Machine | Rough age | Typical settings | Display |
|---|---|---|---|
| Desk one | Assembled 2021, one component replaced since | Medium, capped frame rate | 27-inch, 1440p, 60 Hz |
| Desk two | Assembled 2022 | High where it holds, otherwise medium | 32-inch ultrawide, 75 Hz |
| Laptop | Bought 2023, thin chassis, throttles | Low to medium, always plugged in | Built-in 15-inch, 1080p |
| Living-room box | Assembled 2020, used for controller games | Medium, controller only | Television, 1080p, sat two metres back |
Corrections
When we get something wrong
It happens two or three times a year, usually about performance or a system we misread.
Corrections are appended at the foot of the entry with the date and a short description of what changed. We do not silently rewrite sentences: if a claim was wrong, the wrong sentence stays on the page with the correction beneath it, because quietly editing history makes every other page less trustworthy.
If the error is serious enough to change a verdict label, we say so explicitly in the correction and leave the original label struck through. That has happened once, in 2024, when a patch during our fortnight altered a system one of us had criticised at length.
Spotted something? The talk back page has a form, and corrections go to the front of the reading pile.
Refusals
Four things we do not do
If a second player never materialises, the game does not get an article. We have a folder of forty hours of notes on games nobody else at the desk had time for, and none of it will ever be published, because a solo opinion is exactly the thing this site was built to avoid.
There is no scale on which a split verdict can be compared to a consensus verdict. Inventing one would require pretending the two columns are a single opinion, which would undo the format. So there are no numbers here, no stars, no ordered lists and no annual round-up.
Two people finding forty hours each is slow. Rather than publishing one half of a pair on launch day and the other three weeks later, we hold the whole thing until both columns are in. The result is that our coverage runs weeks or months behind, and we have made peace with that.
Where a game reached us through a studio or a press contact, that is stated at the foot of the entry. It changes nothing about what gets written. Two of the four of us have lost access to a mailing list over a published column, and the columns stayed as filed.