Turn & Banking Gameplay : rotation under constraint
- Vanessa
- 6 hours ago
- 14 min read
The art of cheating in locomotion
You push the stick to the left at full speed.
The character is supposed to turn around.
And in that moment, either the pivot feels disconnected from your input, or it drags and slips, and you immediately sense friction between what you intend and what the screen gives back.
It’s the moment where locomotion reveals whether it can absorb a rotation without losing the player.
The turn is a sensitive building block. It reuses familiar constraints (support foot, angles, blends), but exposes them differently because it must operate at full speed, under constantly shifting analog input, and inside the continuity of the locomotion cycle.
Banking works in another register. It doesn’t change gameplay, but it shapes how the player perceives that gameplay. A light animation layer that says a lot about the character’s credibility in motion: direction, weight, intention.
I treat these two blocks together because they address the same gesture: turning while moving.
The turn handles the actual rotation. Banking handles the perceived rotation.
And together, they show how a game negotiates a movement that is simple for the player, but demanding for the system.
This month, we’re looking at the art of cheating in gameplay: how to bend biomechanics intelligently to preserve gamefeel, why the engine “eats” rotation angles, and why banking is not just an aesthetic flourish.
Download the PDF “Gameplay Animation Framework – TURN & BANKING” (Frameworks section)
5 pages to frame turns before production: a complete Decision Sheet, a TURN transition map, a risk chart, and the list of forbidden states.
The boundary between cycle and rotation
The start and stop have an easier job: they mostly handle an intensity shift from or toward immobility.
The turn doesn’t react to a change in speed. It reacts to a vector.
A direction that evolves while moving: twelve degrees left, then forty‑five, then one‑hundred‑eighty.
And the engine must decide, frame by frame, whether a subtle reorientation of the current cycle is enough… or whether a dedicated rotation animation needs to fire.
So before animating anything, there’s a systemic question to answer: where do we draw the line between a simple directional blend and a true turn?
Because a turn doesn’t have the luxury of starting from zero speed.
It triggers during movement. The character already carries energy, weight, and velocity in one direction, and must change that direction without breaking momentum, without breaking the cycle, without losing continuity.
That’s why a turn immediately exposes the coherence of a locomotion system.
It shows whether the engine, animation, and game design can interpret the same intention at the same moment, under real‑time analog input constraints.
We don’t create a turn to look good in isolation.
We create it to absorb the chaos of the controller. It has to become a physical, real‑time response.
Reactivity vs Credibility
In reality, making a full 180‑degree turn while running is a heavy operation.
The body slows down, plants its support foot, lowers its center of gravity, absorbs inertia, then drives the mass into the new direction.
It’s credible, it’s physical, and it’s far too slow for a player.
If you impose that biomechanical truth as‑is, the rotation becomes sluggish.
The character feels like it weighs three tons, and the input seems to register half a second late.
Real biomechanics are simply… too real.
So we compress. We cheat. We lie.
We shrink twelve frames into three. We start rotating earlier than the body realistically could. We shorten the support phase so the response feels immediate, even if the movement shouldn’t be able to turn yet.
But if we cheat too much, the character loses all physical consistency. It spins like a top, the feet slide, the weight disappears. And the player feels the contract has been broken.
That’s the paradox of the turn: cheat enough to stay reactive, but not so much that you lose credibility.
And that line isn’t found in Maya or in a design document.
It’s negotiated with the controller in hand, directly in the engine, against the noise of real‑time analog input.
The eaten angle
What Maya shows is never exactly what the engine delivers.
On a turn, that discrepancy becomes particularly treacherous.
Here’s the classic scenario: you animate a perfect 180‑degree turn.
You set up the entry blend so the run connects cleanly into the start of the rotation.
But during those few transition frames, the engine gets ahead of you. It already begins rotating the capsule on its own.
Result: when the animation actually starts, part of the rotation has already been consumed. You animated 180°, the engine outputs 150°.
The character ends up facing a direction different from the one you planned.
And that offset, invisible in Maya, becomes disorienting in‑game.
This is the eaten angle phenomenon, one of the hardest bugs to track down, because it exists only inside the engine, at a specific speed, under a precise blend condition.
On Beyond: Two Souls, Quantic Dream’s proprietary engine handled foot synchronization natively.
Each animation carried its own foot data, and the system knew exactly which foot was on the ground at every frame.
This made it possible to produce four turns to cover every situation:
180° left on left foot
180° left on right foot
180° right on left foot
180° right on right foot
When the player gave the input, the engine read the current support foot and triggered the corresponding turn instantly.
Zero latency, zero sliding.
On Ghost Recon, the engine didn’t have this native foot‑reading capability.
You had to compensate through strict entry and exit posing, and through precise root‑motion placement to avoid sliding.
More constraints during production, but the same goal: preventing the engine from devouring what the animation was meant to deliver.
The standard we established on Beyond to contain the damage was:
The animation starts on a passing pose.
You let a full step play out until the next passing pose for the entry blend.
Then the rotation can finally begin. It must end on the following passing pose.
Finally, you add one extra full step for the exit blend.
A protected window where the engine couldn’t bite into the actual degrees of rotation.
It was a compromise. It always is.
Suspension of control
Here’s the real tension inside a turn, the one you never see in tutorials: how many frames will the player tolerate being out of control?
Because that’s exactly what a turn imposes.
It asks the player to briefly let go of the controls while the character pivots, before handing control back once the rotation is complete.
The player is a chaotic creature. At frame 3 of a 180° turn, they spot an enemy and push the stick forward again. The player gives a direction, the character must pivot, and during that short window, control is suspended. It returns only once the rotation is finished.
If this window of non‑control is too long, the system becomes purely blocking. The character feels stuck in its half‑turn like a cargo ship.
The player waits, and it feels like they’ve triggered a micro‑cutscene.
And in an action game, waiting means dying.
If the window is too short, the system tries to give control back too early. The rotation gets truncated, the pivot never completes. The character ties itself in knots, slides two meters, and the entire credibility of the movement collapses.
The rotation must be fully completed before the next input is allowed. This rule is non‑negotiable.
Letting the player interrupt a turn mid‑rotation opens the door to impossible entanglements: the character twisting into improbable angles, the root motion or capsule drifting one way while the locomotion cycle demands another.
On Beyond: Two Souls, that window was protected.
The entry blend played out, the root motion began its rotation, and the system would only accept a new input once the rotation was fully completed, right before the exit step.
A short window, but a defined one. And in the context of a narrative game, that was acceptable.
In a pure action game, it’s a different story.
Rotation is no longer handled by root motion but by the engine capsule.
Every frame of non‑control feels like an eternity.
The pressure to shorten that window is maximal and that’s where the cheating has to be the most precise.
The camera: the other half of the turn
One often underestimated element in a turn: the camera.
A turn can be perfectly animated and perfectly integrated, yet still disorient the player.
Because a turn doesn’t live alone. It lives with the camera ,and the camera directly shapes how the movement is perceived.
In 3D, a 180° turn without a coherent camera is guaranteed disorientation.
The character pivots, the space flips, and if the camera doesn’t follow the movement with the right logic, the player loses their anchor in the environment. They’re controlling something, but no longer know what or in which direction.
On Heavy Rain, camera angles changed every time you entered a new room. It was a cinematic narrative game: the camera served the story, not the controls. The turns were fine, but the camera axis could shift at the exact moment the character changed direction.
It wasn’t a camera problem, nor an animation problem, it was a 3C alignment problem.
The camera served the narrative.
The animation served the movement.
The player needed spatial continuity.
When two systems pursue different goals, sensations crack, even if each system does its job perfectly.
It’s a structural contradiction.
And in a video game, even a cinematic one, everything eventually converges: if the camera doesn’t serve control, the player loses the thread.
In that context, animation had to absorb the tension. Because it was the only flexible system in an architecture where the camera had a narrative mission that took priority.
On Beyond: Two Souls, we knew this issue was inherited from Heavy Rain.
We worked our turns differently, reducing the duration of the rotation window and producing cleaner animations on the transitions.
But the collaboration between camera and animation hadn’t really evolved.
In a narrative game, the camera helps tell the story and the characters, and it’s often the animation that absorbs the consequences.
A turn is a collaborative decision.
Animation handles the feet, the root motion, the duration of the rotation.
The camera handles readability, narration, and the world axis. If one of them fails to hold its part, the other collapses.
You can cheat on frames, hide foot contacts, or lock the rotation, but you can’t animate in place of the player’s eye.
In the end, the perfect movement isn’t the one that looks good in Maya: it’s the one the camera allows the player to understand.
Two ways of thinking about a turn
In production, there are two major ways to think about a turn.
These aren’t animator choices, they’re gameplay trade‑offs, control philosophies.
The “fluid curve” approach
This is the approach used in narrative games and realistic open worlds.
Root motion drives the trajectory. The character follows a clean curve, and the animation controls every degree of rotation. The result is credible, organic, coherent.
This is the approach of Heavy Rain, Beyond Two Souls, Assassin’s Creed, Ghost Recon. Animation serves credibility, and reactivity is cheated just enough to stay within the plausible.
You never reproduce reality: you compress it, simplify it, arrange it but you try to stay in a zone where the player believes what they see.
The “brutal cheat” approach
This is the approach of pure action games.
The character capsule snaps instantly toward the requested direction, and the engine handles the actual trajectory.
Animation only dresses the result. The feet may slide for a frame, the hips may lag slightly but the player doesn’t notice any of it.
What they feel is a character that responds instantly, effortlessly.
This is what I discovered on Prince of Persia: The Lost Crown.
And it didn’t just change the way I think about turns, it changed the way I think about gameplay itself.
All the animations. All the transitions.
The entire logic of how a character should respond to the player.
At the start of production, I resisted. Letting the engine handle trajectory without root motion felt like opening the door to sliding, to visual breaks, to a loss of quality and I cared deeply about that.
I defended my arguments. The game director listened and proposed a compromise: some animations in root motion, others driven by the engine. I wasn’t convinced at first.
I was wrong. The in‑game sensations were excellent, and the visual quality hadn’t been sacrificed on the altar of responsiveness.
I learned something fundamental: a well‑measured cheat is often better than the truth.
Today, that shift has changed my perspective.
Even on realistic productions, we can go much further than what we allowed ourselves back then.
The boundary between absolute respect for the body and systemic cheating isn’t a red line. It’s a slider and you have to dare to move it.
It’s not “root motion or engine.” It’s not “credibility or responsiveness.”
It’s a simple question: what sensation do we want the player to feel?
Banking: visual sound design
Banking is often misunderstood in locomotion.
It’s frequently treated as an aesthetic detail, a slight body tilt in turns, something to make the movement “prettier,” to give a bit of life to the walk cycle.
It’s a category error.
Banking isn’t aesthetics. It’s psychology.
Here’s what actually happens in the engine: when the character turns, its capsule stays perfectly vertical. It doesn’t lean. It doesn’t experience any centrifugal force. It simply pivots on its axis, cleanly, without inertia.
The player, however, expects to feel the movement.
A mass that resists. A weight that leans into the turn. A force pulling the character outward along the curve.
That force doesn’t exist in the engine. So the animation creates it.
There are two major ways to produce banking:
Additive layer : It tilts the spine, shifts the pelvis slightly toward the inside of the turn, and orients the head and shoulders. This version does not affect the actual trajectory: it doesn’t touch root motion.
Inclined cycle (left/right) blended : The inclined cycle can be animated with root motion or in place.
With root motion, the cycle carries a slight rotation or displacement, which can influence the real trajectory. This isn’t the goal of banking, just a possible consequence of the pipeline.
In place, the capsule handles all movement and orientation: in that case, banking has no impact on trajectory.
In both approaches, the intention remains the same: injecting a psychological sensation, not a physical mechanic.
All of this tells the player’s brain: this character is experiencing the inertia of its own speed.
This is exactly what sound design does: creating sensory information that doesn’t exist in the physical reality of the system, so the player feels something true.
On Beyond: Two Souls, we searched for the right banking angle in mocap.
We asked actors to walk while turning, in circles that grew tighter and tighter, down to the smallest possible radius.
That gave us every angle, every intensity.
Then we tested in‑engine to find what looked best depending on gameplay reactivity.
Because the correct level of banking depends directly on how quickly the stick influences the trajectory.
A hyper‑reactive gameplay where a tiny flick of the stick triggers a 90° turn requires banking that is radically different from a gameplay where the trajectory is more progressive.
What banking never does : It never controls gameplay. It never decides trajectory, speed, or reactivity.
Even when blended with root motion, its role is not to drive movement, its role is to tell the story of that movement.
It is a purely visual layer. And that’s exactly why it must be designed separately from the turn but in absolute coherence with it.
Turn + banking : real control vs perceived control
Taken separately, turn and banking serve different objectives.
The turn: real control
The turn is mechanical. It manages trajectory, rotation, the control window, foot transitions. It has a direct impact on what the player can do, and when they can do it. It’s the system that guarantees the player’s input becomes an actual movement in the world.
The banking: perceived control
Banking isn’t aesthetic it’s perceptual. It dresses the rotation, gives weight, personality, credible inertia. It isn’t meant to control what the player can do, but it deeply influences what the player believes they are doing.
Banking tells the story of the physics the engine doesn’t have. Turn tells the story of the physics the engine does have.
Together: the sensation of mastery
Together, they form something greater: the boundary between real control and perceived control.
The turn ensures the player truly controls the character’s trajectory.
Banking ensures that this control feels physical, credible, alive.
When both are properly balanced, the player sees nothing.
They don’t think about the animation. They don’t think about the system. They think about their goal in the game and that is exactly what good gameplay locomotion should produce.
NPC vs joueur : two logics, two constraints
A final distinction that changes everything in production: player‑controlled turns and NPC turns do not obey the same constraints.
The player: the reign of chaos
The player is unpredictable. They change their mind three frames into a twenty‑frame turn. They send chaotic, rapid, contradictory inputs.
Yet they must remain in control at every instant or at least feel absolutely in control.
For the player, turns must be cleanly interruptible, compatible with every foot support, and minimize the non‑control window as much as possible.
The NPC: the power of prediction
The NPC, on the other hand, can predict. The AI knows where it’s going and knows the obstacles ahead. It can deliberately choose between a progressive turn with banking to follow a wall, or a sharp 180° to react to a threat.
The AI can optimize the trajectory before the first frame of animation even begins.
But this superpower has a limit: the unexpected. If a dynamic object or another character cuts across the NPC’s path in the middle of its rotation, the prediction collapses.
The AI suddenly becomes as chaotic as a player. The system has no choice: it must cheat. It cuts the planned turn to force an emergency stop or an avoidance stumble. The animation absorbs the shock, the feet slide for a frame, and visual credibility temporarily fades to give priority to collision handling.
Because outside its ideal trajectory, the NPC becomes subject again to the hard laws of the engine.
This difference changes production decisions.
Player turns must be cleanly interruptible, compatible with all foot supports, and minimize the non‑control window.
NPC turns can be longer, more varied, more nuanced — because the AI triggering them already knows what comes next.
Why some games make turns more critical than others
The criticality of turns depends directly on the type of space the game takes place in.
In a game with narrow, realistic, human‑scale environments — like Beyond: Two Souls, every rotation becomes visible, noticeable, sometimes risky for readability. Mistakes don’t forgive.
Conversely, in a game with wide, open environments, or one where the player spends most of their time strafing or in cover, like Ghost Recon, turns in free locomotion become mechanically less critical.
It’s a matter of gameplay context.
Put your production effort where the player is actually looking.
Conclusion
There’s one thing I wish I had understood earlier in my career.
For years, I worked on systems where animation drove control. Root motion defined the trajectory; animation was the source of truth.
I defended this approach because it produced credible, controlled results.
Prince of Persia: The Lost Crown taught me something else.
That a well‑measured cheat is often better than the truth. That players don’t judge biomechanics, they judge the sensation of control.
Animation isn’t there to reproduce reality. It’s there to build an illusion solid enough for the player to forget they’re holding a controller.
What turns and banking reveal isn’t just a story of rotation.
It’s a story of control, perception, and sensation.
The turn tells the engine where to go. Banking tells the player what they feel.
Over time, I understood that nothing is “true” or “false.” There isn’t a good method and a bad one. There are philosophies, contexts, gamefeel intentions.
There are unpredictable players, predictive NPCs, engines that cheat, and spaces that amplify problems.
In the end, only one question keeps coming back: what should the player feel?
Everything happens in that fragile boundary between what the engine actually does and what the player believes they’re doing.
If that boundary disappears, if everything aligns, the player stops thinking about the system and focuses on their direction.
A successful turn isn’t a realistic turn. It’s an invisible one.
A gameplay animator doesn’t animate rotation, they seal a responsiveness contract with the player’s stick, frame by frame.
And that is exactly what good locomotion should produce.
And if you want to go further, check out the Turns & Banking Challenge — July 2026 and the Locomotion Foundations Kit.
Download the PDF “Gameplay Animation Framework – TURN & BANKING” (Frameworks section)
5 pages to frame turns before production: a complete Decision Sheet, a TURN transition map, a risk chart, and the list of forbidden states.
Comments