The Floor Is the Hardest Problem in Games
You drop from a ledge. The character hits the ground. Something is wrong.
Not broken, exactly. Just hollow, like watching a cardboard box fall instead of a person. You've felt this in bad games without being able to name it, and the inverse is true too: you've felt a landing in a well-made game and had a brief, wordless sense of yes, that's right, before moving on without ever stopping to ask why.
That feeling isn't accidental. It's four or five overlapping systems firing in the same fraction of a second, each doing a specific mechanical job. Pull any one of them out and the illusion collapses.
Velocity Is Memory
The simulation of weight starts before the character ever leaves the ground. Every physics-capable engine tracks a falling body's velocity on the vertical axis, and that number accumulates across the entire fall. A character who drops two meters arrives at the ground carrying a very different value than one who drops eight.
That stored velocity is the engine's memory of how far the fall was.
The simplest use is the landing animation selector. Unreal Engine lets designers set thresholds: below a certain fall velocity, play the light-step animation; above the next threshold, play the heavy-land; above the highest threshold, play the stagger or the ragdoll. A one-meter hop and a four-meter drop produce visibly different responses because they arrived with genuinely different physics histories. The number isn't cosmetic. It drove the decision.
What makes this feel like weight rather than just animation variety is the blending. A good implementation doesn't snap between states. It interpolates, so a fall that lands exactly on a threshold produces a blend of the two animations on either side, weighted by proximity to each cutoff. The seam disappears.
The Squash That Your Eye Expects
Human perception is surprisingly well-calibrated to squash-and-stretch, the animator's oldest principle. When something heavy lands, it compresses briefly before rebounding. Your visual cortex has been watching this happen since infancy, and it will notice the absence before you consciously register anything wrong.
Game engines handle this two ways, and the better ones use both.
First: procedural bone scaling. The spine, the knees, the ankles get briefly scaled along the vertical axis at the moment of impact, compressing the character's silhouette by a few percent for a fraction of a second before releasing. In games built on Unity's Humanoid rig system, this procedural deformation layers on top of a base animation without replacing it, so a designer can dial the intensity without re-animating anything.
Second, and this one is underrated: camera response. A slight downward push on the camera, followed by a dampened spring-back, tells your brain that something with mass just made contact. Think of it like a hand holding a phone while someone nearby jumps off a table. The camera felt the impact. The spring physics are tuned so the initial drop is fast and the recovery is slow, which matches how real spring systems behave and feels distinctly heavier than a symmetric oscillation.
Take both away and the character looks like it's landing on a trampoline made of air.
The Sound Doing Half the Work
Here's something the graphics side of the industry consistently undersells, and I'll say it plainly: audio is doing roughly half the perceptual work in a landing, and pretending otherwise leads to bad design decisions.
Picture a character dropping from a rooftop in a game with excellent animation but placeholder sound, a generic thud with no low-end. Now the same drop, same animation, but with a layered impact: a sharp transient crack, a subsonic thump sitting around 60-80Hz, and a brief reverb tail that shifts depending on whether the floor is concrete, wood, or dirt. The second landing will feel heavier to almost every player, even those who couldn't articulate why.
Modern engines use physical material tags to drive this. Unreal's Chaos physics system and Unity's PhysX layer both support surface type data attached to geometry, so when a character's foot collides with a tagged surface, the audio system queries that tag and selects from the correct sound pool. A landing on gravel and a landing on metal trigger different audio events automatically, no hard-coding required.
The subsonic element is what most players consciously miss but unconsciously rely on. Very low frequencies are felt in the chest as much as heard in the ears. They signal mass. A landing with no low-end content feels light regardless of what the animation does.
The One System That Breaks If You Rush It
Inverse kinematics, IK for short, is where landing weight simulation gets genuinely complicated.
The problem is terrain. A character landing on a flat test floor is easy. A character landing on a slope, a rock, or a pile of rubble is a different problem entirely. Without IK, the animation system plays the flat-ground landing animation regardless of the actual geometry. The character's feet float above a slope, or clip through it.
IK solves this by working backwards from the foot contact point. The engine plants the foot where the ground actually is, then resolves up the kinematic chain: ankle, knee, hip, spine. Each joint adjusts procedurally to accommodate the real contact point without breaking the skeleton's proportions. Horizon Forbidden West's landing system is a well-documented example of this done at high fidelity, with Aloy's foot placement adapting in real time to terrain no specific animation ever anticipated.
The reason this breaks when rushed: IK chains need constraints. Without careful limits on joint rotation, the solver finds mathematically valid solutions that are anatomically horrifying. A knee bending the wrong way is technically a correct IK solution. It is also immediately, viscerally wrong to any human observer. Tuning those constraints is painstaking work, and largely invisible when done right.
This is where the gap between a large studio and a small one shows up most clearly, not because small teams lack the knowledge, but because tuning IK properly takes time that small budgets often can't absorb.
What People Consistently Misread
There's a widespread assumption, even among players who think carefully about this, that heavier-feeling landings come from slower animations. Slow the playback down and it'll feel more weighty. This is wrong, and it produces some of the worst-feeling landings in the medium.
Weight isn't slowness. Weight is resistance and consequence.
A heavy landing can be fast. The initial impact is fast. What makes it feel heavy is the squash on contact, the camera lag, the low-frequency audio hit, and then the slightly slower recovery: the grunt, the moment before the character is ready to move again. The impact is a fast event followed by slow consequences. Slow the impact itself and you get something that reads as sluggish, not massive.
And here's something that deserves more attention than it usually gets: two players who bought the same action game can have completely different experiences of whether it feels good, based entirely on whether they're playing on speakers with subwoofer output or laptop speakers with nothing below 150Hz. The physics and animation are identical. The perception isn't. Audio hardware is doing a surprising amount of the heavy lifting, which means a game's landing design is only ever as complete as the weakest speaker in the room it's played on.
The Number That Matters
If you want a practical lens for evaluating any of this: find a consistent drop in a game you're playing, maybe off a specific ledge, and do it twice. Once while watching the character. Once while listening with your eyes closed.
If the second version feels less convincing than the first, the audio team didn't finish the job. If the first version feels less convincing than the second, the animation is leaning on sound to carry weight it should be generating itself.
The best implementations make both versions feel right, because they're not compensating for each other. Five systems, all firing at once, all agreeing on the same answer. That's not polish. That's the whole architecture.