How Robot Games Use Physics

Arcade physics is not a simplified version of real physics. It is a different thing built for a different purpose: being predictable. This guide covers what that means in practice and why the games feel better for it.

Predictability over accuracy

Real physics has variable air resistance, spin, friction that depends on surface and speed, and a dozen other factors. All of them make outcomes harder to predict from a glance, which is exactly wrong for a game where you have a quarter of a second to decide.

Arcade physics strips this down to constant gravity, a simple drag term and nothing else. The result is that a jump looks the same every time, a thrown object follows the same arc every time, and a player can learn the shape after two or three observations.

That predictability is the whole point. A game where the same input produces the same result is one where failure is informative; a game with realistic variability is one where failure is ambiguous.

The fixed jump arc

In most arcade platformers the jump is entirely fixed: same height, same distance, same hang time, regardless of how long you hold the key.

This is frequently criticised as less expressive than a variable-height jump, and for some games that criticism is right. But it buys something specific and valuable: every failed jump is diagnosable. If you fell short, you launched from the wrong place, because the arc cannot have been the problem.

Variable jumps make failure ambiguous - did you hold too briefly, or start too late? Two possible causes means twice as long to learn, and in a game with runs measured in minutes that matters.

Coyote time and jump buffering

Two adjustments almost every good platformer includes and almost none explain.

Coyote time is a short window after you walk off a ledge during which a jump still works - typically around a tenth of a second. Without it, a player who presses jump a frame or two after leaving the platform simply falls, which feels like the game ignoring an input.

Jump buffering is the reverse: a jump pressed slightly before you land is remembered and fires on touchdown. Without it, chaining a landing into an immediate jump requires frame-perfect timing for no reason.

Neither makes a game easier in any visible way. They exist so that failures are decisions rather than timing margins no human can reliably hit, and their absence is felt as the game being unresponsive rather than as being hard.

Momentum and why it is usually limited

Most arcade games give the player very little momentum - you start and stop almost immediately - because momentum makes precise positioning difficult and precise positioning is usually the skill being tested.

The exceptions are deliberate. Mech games add momentum specifically to make position a planning problem. Low-gravity games add it because drifting is the mechanic. In both cases the momentum is the point rather than an attempt at realism.

Where momentum exists, the design has to compensate elsewhere: wider platforms, longer telegraphs, more forgiving hazard margins. A game that adds momentum without adjusting anything else is simply harder to control, which is not the same as being deeper.

Low gravity and coasting

Space-themed games often reduce gravity, which changes the economics of movement rather than just its feel.

With low gravity and light drag, a short burst of thrust followed by a long coast covers nearly as much ground as continuous thrust and costs a fraction as much fuel. That turns efficient movement into a rhythm - burn, coast, correct - rather than a matter of holding a direction.

The counterintuitive consequence is that arriving slowly is worth more than arriving quickly, because stopping also costs thrust. Players who hold the throttle for whole crossings run out of fuel in the same place every time.

Tuning a jump, in practice

A concrete look at how a jump arc gets chosen, since it determines most of how a platformer feels.

The two numbers are gravity and initial upward velocity, and they are not chosen independently. What gets chosen first is the feel: how high the jump should reach, and how long it should last. Those are the things a player perceives.

Height and duration then determine the other two. A jump that peaks at roughly two and a half times the robot's height and lasts around three quarters of a second feels responsive without being floaty; longer hang times read as low gravity and start to feel like swimming.

The horizontal distance follows from duration and movement speed together, which is why changing a character's run speed silently changes every jump in the game. That coupling is the usual reason a platformer feels inconsistent after a speed tweak.

Arcade platformers also tend to use higher gravity than realism would suggest, with a correspondingly higher launch velocity. It produces a snappier arc that reaches its peak quickly and falls decisively, which is easier to judge than a slow symmetric parabola.

Knockback as a design tool

Knockback is usually thought of as feedback and is doing real mechanical work.

It creates space. A hit that pushes an enemy back buys you the distance to follow up or disengage, which means landing a hit is rewarded twice - once in damage and once in position.

It also controls pacing. Heavy attacks push further than light ones, which means committing to a heavy attack resets the engagement distance and gives both sides a moment. Without it, a fight collapses into permanent contact, which is far less readable.

And it communicates weight. A machine that barely moves when hit reads as heavy regardless of its size, which is a cheaper and more reliable signal than the visual alone.

The design constraint is that knockback must be predictable. Random variation in how far something is pushed makes follow-ups impossible to plan, which removes most of the benefit.

Frequently asked questions