07Quarterly
Weight And Momentum
How movement feels, measured
Weight And Momentum
How movement feels, measured
Games almost never tell you how large anything is, so every spatial figure we publish rests on a scale we established ourselves. This note explains how, and why an error here is worse than an error anywhere else in the process.
Applies to driving games, mechs and any figure quoted in metres · Requires a free or detachable camera
A frame count is self-contained: you either counted correctly or you did not, and an error affects one figure. A distance is not. Every stopping distance, turning circle and displacement we publish is derived from a scale, and if the scale is wrong then every spatial figure in the review is wrong by the same proportion, consistently, in a way that looks entirely plausible. We know because it happened to us in issue five.
The reliable reference is not the vehicle and not the environment. It is a human figure, if the game contains one, because human height is the one dimension a game's art team will have modelled against a real value. We take a standing character, measure their height in screen pixels with the camera at a fixed known position, and assume 1.75 metres.
Where no human figure exists — the mech game in this issue has none at usable scale — the fallback is architectural: door height, step rise, or handrail height, in that order of preference. Doors are modelled at approximately two metres in essentially every game we have measured. Steps are more variable but cluster tightly around 0.17 metres, and a staircase gives you many of them to average across.
We established scale from a vehicle's stated wheelbase, taken from an in-game specification screen. The specification was in the game's fictional units, which did not correspond to metres, and every distance in that review came out roughly 1.3 times too large.
Two readers repeated the measurement and reported the discrepancy independently. The correction is noted at the top of that review, and the rule since then is that scale never comes from anything the game states — only from something it draws.
Never derive a scale from one thing. The procedure is to establish it twice, from unrelated references, and accept it only if the two agree within five per cent. A human figure and a doorway are independent enough for this purpose; a doorway and a corridor width are not, because both come from the same architectural kit and may share the same error.
| Reference | Assumed value | Observed reliability |
|---|---|---|
| Standing human figure | 1.75m | High |
| Doorway height | 2.00m | High |
| Step rise | 0.17m | Moderate |
| Handrail height | 1.00m | Moderate |
| Road lane width | 3.50m | Low |
| Anything the game states | — | Unusable |
Lane width is on the list because it is tempting and because it is unreliable: games widen roads for camera reasons far more often than they widen doorways, and a lane measured at four and a half metres tells you about the level designer rather than about the world.
With a scale fixed, the turning circle itself is straightforward. Hold full lock at a constant speed, capture from directly overhead if the game permits a free camera, and measure the diameter of the path traced. Where an overhead view is unavailable, drive the circle beside a straight reference — a wall, a painted line — and measure the maximum lateral displacement, which is the radius.
Three details matter. Measure at constant speed, because turning circles in most games vary with velocity and a decelerating circle is a spiral. Measure a full revolution rather than a quarter, because the first quarter includes turn-in transient. And measure both directions: three of the eighteen games we have measured across seven issues had asymmetric turning circles, in two cases by more than eight per cent, which is almost certainly unintended.
Any figure in metres that we publish rests on a doorway or a human figure. If those are wrong, everything spatial in the review is wrong together, which is the hardest kind of error to notice.
We publish distances to two significant figures and nothing more. The precision available from this method is roughly five per cent, and printing a stopping distance as 31.4 metres implies an accuracy the technique does not have. Thirty-one metres is honest; 31.4 is decoration.
Where a figure matters to an argument — as the four-fold mass discrepancy does in review three — we state the scale reference used and the second reference it was checked against, so that a reader who disputes the conclusion can dispute the foundation rather than guessing at it.