07Quarterly
Weight And Momentum
How movement feels, measured
Weight And Momentum
How movement feels, measured
Nothing about this character obeys physics. Acceleration is two frames, deceleration is zero, air control is total, and at the apex of a jump horizontal momentum is simply discarded. It is the most artificial movement system in the issue and one of the two best.
19 hours · 1,940 deaths logged · Capture at 240fps, keyboard and controller both measured
The temptation with a publication like this one is to reward simulation. We measure mass, so games that model mass carefully ought to score well and games that ignore it ought not to. This game exists to demonstrate that the temptation is wrong, and we have put it second in the issue rather than fifth for exactly that reason.
Consider what the numbers say. The character reaches full running speed in two frames. Release the stick and speed is zero on the next frame, with no skid, no lean and no decay. In the air the player has complete authority over horizontal position at full ground speed, and at the top of a jump the game deletes horizontal momentum entirely so that the descent is a controlled drop rather than a continuation of an arc. A physicist would call this nonsense. A speedrunner would call it a language.
| Property | Value | Notes |
|---|---|---|
| Standing to full run | 2f | Linear, no curve |
| Full run to standstill | 0f | Instant |
| Direction change | 1f | Full speed retained |
| Jump rise | 17f | Height varies with hold |
| Apex hold | 2f | Momentum discarded after |
| Jump fall | 14f | Faster than rise |
| Coyote window | 5f | Jump accepted after leaving ledge |
| Input buffer | 6f | Jump accepted before landing |
The two rows at the bottom are the reason the game feels fair while being brutally difficult. A five-frame coyote window and a six-frame input buffer mean that eleven frames of the player's imprecision are absorbed silently, and in nineteen hours neither of us ever felt cheated by a jump. We died one thousand nine hundred and forty times and blamed ourselves on essentially every occasion.
The asymmetric arc — seventeen frames up, fourteen down — is standard practice and worth explaining because it is so widely used and so rarely stated. A faster fall than rise makes the peak of the jump feel like a decision point rather than a plateau, and shortens the interval during which the player has committed and cannot act. In a game with hazards below, it is the difference between a jump and a gamble.
The apex behaviour is the design decision we expected to dislike and now consider the game's best idea. Because horizontal speed vanishes at the top of the arc, a jump's landing point is a function of the player's input during the descent rather than the run-up. That converts every jump from a ballistic problem into a positional one, and it means the hardest sequences in the game are tests of reading rather than of committing.
Compare review one, where four frames of uncancellable commitment is the whole point. Both games are about decisions; one charges you for changing your mind and the other refuses to. The reason both work is that each is consistent with what its sequences ask. This game's sequences are dense, fast and require constant revision, and a character with mass would make revision impossible.
Realism is not the standard and never has been. The standard is whether the movement can express what the level demands.
Everything the game gains in precision it loses in physicality, and there are moments where the absence is felt. Landing from a great height looks identical to stepping off a box: no recovery, no dust, no impact. In a game whose art suggests a real body moving through real stone, that flattening is a genuine loss, and it is most obvious in the few scripted moments where the game wants the player to feel small.
Moving platforms do impart momentum, which nothing else in the game does. Jump from a platform travelling laterally and the character carries its velocity into the arc — the only case in nineteen hours where inertia exists.
It is almost certainly an engine default rather than a decision, and it produces the game's only unfair-feeling deaths. Six of our 1,940 were this, and all six felt like the game breaking its own rules.
The other cost is that the game is unreadable to a newcomer for about ten minutes. Because the character has no acceleration tell and no skid, there is nothing to watch to learn what the movement will do; you simply try it. Games with mass teach themselves visually. This one has to be taught by failure, which it does efficiently and without sympathy.
We measured both because with two-frame acceleration the input device is a larger factor than usual. Keyboard is marginally better in this game and the reason is structural: with instant stops and full air control, analogue nuance is worthless, and a digital input that reaches full deflection immediately loses nothing. On a controller the stick's own travel time adds between one and three frames that the game cannot remove.
That is the only game in this issue where we would advise a specific input device, and the advice is mild. Three frames on a five-frame coyote window is not nothing, but the game's generosity absorbs most of it.