Write to the bench

Recounted figures

Every number published here was counted by hand from capture taken at 240 frames per second: acceleration ramps, deceleration tails, the frames of commitment a direction change costs. Two figures across seven issues have proved wrong, both found by readers who repeated the measurement, and one of them changed a review's conclusion. That is the single most valuable thing anyone reading this can do.

For a recount to be usable it needs the conditions: the game build, the platform, the frame rate the game was running at, and whether any assistance or accessibility option that alters movement was enabled. A game that accelerates over eleven frames at sixty frames per second and over twenty-two at a hundred and twenty is not two different measurements — it is one measurement described badly, and if we described it badly we want to know.

Captures from other hardware

The bench runs on one display and one controller, which the method page states plainly as a limitation. Readers who have captured the same movement on different hardware — a different frame cap, a variable refresh display, a controller with a different polling rate — routinely find behaviour we could not have seen, and the second of this issue's bench notes exists because of exactly such a message.

Raw capture files are welcome; so is a clear description with the frame numbers marked. Both are credited however you prefer, including not at all.

Disagreement about design

Whether nineteen frames of deceleration is generous or merely slow is a judgement, and readers who think the lead review gets it wrong are disagreeing rather than correcting. Those letters are read closely and often decide what gets measured next. They do not produce edits, and the three of us disagree with each other about the same question anyway.

Not answered

Review-copy offers, early access and embargo arrangements, because every game here is bought at retail and played for at least fifteen hours before anything is written, and a free key cannot change either condition. Hardware supplied for testing, for the same reason: the bench buys what it measures so it can say what it likes about it.

Deleted after the first line: sponsored posts, affiliate schemes, link exchanges, and offers to improve the site's search performance.

Reply times

Recounts and corrections: a few days. Captures: a week or two, since verifying one properly means re-measuring at this end. General letters: read always, answered when there is something worth saying, and the silence implies nothing about the letter.

What happens to your message

It is read, answered where appropriate, then deleted along with the address it arrived from. Nothing is added to a list, nothing shared, nothing joined to your reading here — which no page on this site could manage in any case, since none of them loads a tracker or any third-party file. The privacy page sets out how to confirm that in a browser rather than take it on faith.