What is game feel

Why does jumping in Celeste feel great while jumping in a cheap platformer feels floaty? That’s game feel: doing an action and getting the result in the same moment. The loop already asks for that result. Feel is how the result lands in the hands: whether it arrives when the player expects, whether the space gives it something to meet, and whether the picture and the sound sell the meeting. It cannot tell you the dodge landed. That is why AI tools sit next to Play mode, not in place of it. If you do not have a Unity project yet, how to make a game in Unity is the first Play button.

Steve Swink’s book Game Feel (2009) puts the same idea in one line. Real-time control of virtual objects in a simulated space, with interactions emphasized by polish. Martin Pichlmair’s entry in the Encyclopedia of Ludic Terms, first published April 21, 2022, starts one step wider. Game feel is the player’s affective reaction to a game situation, how the game feels to be played. The noun implies a person and a context. Two players can meet the same rule and report two sensations. Pichlmair’s point is that you still design for it. The feeling is the sum of decisions you can name.

Swink had already tried to name it in a Game Developer article on November 23, 2007, “Game Feel: The Secret Ingredient.” He calls it slippery and mostly subconscious: sights, sounds, and an instant response to action. If it is off by a little, the game is in trouble. If players call it responsive, tight, and deep, he treats that as the magical case, and he is careful that those words are the player’s words. His example is steering Mario in Super Mario 64, especially in a stage with little pressure, where the pleasure is running, bouncing, and spinning. Control goes in as fast as he can think, and feedback comes back as fast. His sketch of where the hours go is a sketch, not a measured study: the long work of finishing the game, and underneath it the thing you do every moment, which is steering. The sensation of that steering is the foundation the rest of the game sits on.

Three pieces, then the person

The book builds the sensation from three pieces.

Real-time control is a closed loop. The player has an intent, the input carries it, the game reconciles it with its model, and the player sees the change and forms the next intent. Chris Crawford has compared interactivity to a conversation, two agents taking turns listening, thinking, and speaking. Swink says the car is the closer picture for this kind of control. Turning the wheel is action more than thought. Small corrections happen below a conscious plan, and the result feels as if it arrives in the moment it was asked for. His starting definition is precise, continuous control over a moving object. Real-time control of virtual objects.

That definition still needs a space. A ball in a field of blank white has no way to show that it is moving, and nothing to collide with. Simulated space, in the book, is physical interaction the player perceives by moving through it: collision, and the placement of objects relative to how fast the object moves. Those objects give the motion a meaning. Speed needs a frame. A wall is what makes a wall-kick possible.

Polish is the third piece. In the 2007 article he defines it as the impression of physicality that appears when animation, sound, and effects agree with motion that is driven by input. A later chapter of the book separates polish from the simulation itself. The simulation is the part that says two solid things should rebound. Polish is a cue layered on that event. Take the cue away and the rebound still happens. What changes is the player’s picture of how heavy, sharp, or soft the thing was. He is explicit that polish sells the interaction.

Pichlmair lists what else joins those three. Graphics and animation decide the look. Sound, effects, and vibration add weight. Gravity and the other standing parameters of the simulation color every moment. Timing and narrative can intensify a moment. A little forgiveness of a slight mistake can keep a crucial moment from becoming only frustration. Feel is those parts acting together, on a particular player, in a particular moment.

The book also names five experiences a good-feeling game can convey. The aesthetic sensation of control. The pleasure of learning, practicing, and mastering a skill. An extension of the senses. An extension of identity. Interaction with a physical reality that belongs to that game. Pichlmair’s reading is that the list covers the moment and the longer relationship, and that it still describes a wide range of video games. He also says the list leaves games out. A game without real-time control in a simulated physical space does not fit the criteria cleanly. Board games, card games, text adventures, and sports still produce an affective reaction. They have a feel. Swink’s line is the avatar case. The wider term is the reaction. On a first game, both point at the same test: the verb answers, and a stranger can tell.

Where the sensation is missing

What is the moment missing?

The gapThe response is lateNothing to meetThe juice is already loud

The feel is the response in the same moment. Particles on a late action still feel late.

Input, and the device you named

The 2007 article splits a feel prototype into six pieces so the sensation does not get tangled up with the whole design. The first is input: how the player expresses intent. He treats the device as part of the feel. A controller that sits well in the hand changes the session, and a keyboard does not feel like that controller. From prototypes around then, the same motion felt better to him on a wired Xbox 360 pad than on keyboard keys, and he records a complaint that the PS3 pad of that era felt light. That is a 2007 note about those devices. You often do not choose the hardware. You do name the device the row is for, which is already the job of the control scheme.

He borrows Donald Norman’s natural mapping. On a stove, the dials that sit in the same arrangement as the burners need no legend. Geometry Wars, on the Xbox 360 pad, is his game example: the stick moves in a circular gate, and the motion on screen is the same kind of circle. Mario 64 has the related property that rotating the stick tracks the way Mario turns. No instruction has to carry that relationship. The position of the input and the position of the object agree.

Sensitivity is the other half of the device. A button has two states. A mouse can travel across two axes for as far as the desk allows, so it has a huge number of states. Swink calls that input sensitivity, a rough measure of how much expression the device already contains. More sensitivity usually means more complexity, and he refuses to treat that as automatically better. The choice has to match the motion you want. The device is only half. The other half is how the game answers.

Response

Response is how the system processes the input and answers in real time. He compares two games to show that sensitivity can be added or removed in the mapping. Zuma clamps a free mouse down to a rotation in place: the frog looks at the cursor, and the mouse’s freedom is reduced to something the player can manage. Strange Attractors uses a single button, while the ship keeps moving, and the button turns gravity wells on and off. Both, in his account, have a nuanced reaction. He calls the result reaction sensitivity. The core of the feel, he writes, sits in the space between the player and that reaction.

Super Mario Bros. is his proof that a simple device can still answer with a lot of expression. The NES pad is on-or-off buttons. Mario speeds up from rest and slows back down. Holding jump longer makes a higher jump, which is sensitivity across time. Jump plus a direction is a different result from either button alone. Pressing a direction on the ground means something different from pressing it in the air, and the player has to be able to tell which state they are in. Donkey Kong, with a joystick that is a more sensitive device, does not have that fluid answer. A crude input can still be given a sensitive reaction, and a sensitive input can still be muted. There is no formula for the amount. If the motion surprises you and feels like a skill, it is ready for a space.

Context

Context is how constraints give the motion a spatial meaning. Mario in a white void can have a long jump, a triple jump, and a wall-kick on the list, and none of them matter until there is a wall, a gap, or a thing to pick up. Swink’s phrase is that constraints define sensation. Pack objects tightly relative to the avatar’s motion and the space feels clumsy and oppressive. Spread them far apart and tuning stops mattering, and the joy goes flat. He says many massively multiplayer games suffer from that second case to some degree. That is his observation, offered as a spacing warning, not a rule about every online world.

This is why he wants a test environment while the variables are still being tuned. He calls it a magic garden. If the basic interaction is already a pleasure, the rest of the game has a foundation. The historical note in the article is Goddard on Super Mario 64: before the levels existed, Miyamoto had Mario running and picking objects up in a small garden he used to test elements, and a lot of the animation was already there. Swimming took longer to feel right than running. They wanted the water to be an advantage, and fun to dive into, rather than a place the player avoids.

For a prototype, put platforms and some topology in, because a white floor cannot tell you how high a jump should be. A 2D standard unit might be how many tiles a good jump covers. A racing unit might be the width of the road and whether a curve can be taken. His favorite version is primitives in many sizes, spaced against the avatar, then opened up so accidents stay visible. Constraint is what makes a skill. A football field is his analogy: the lines are arbitrary, and without them the skill would be a different one. When a prototype is going well, he notices small goals appearing: through a gap, over a shape, along a line.

Polish, metaphor, and rules

Polish joins the garden around the same time as the space, and only the part the prototype needs. Particles where things meet, a shake of the view, a change of angle, squash and stretch on a collision. The job is to convey physical properties through motion. An effect that supports one consistent physics is in bounds. He tells you to look at real collisions. A hammer on a glass table is noise, motion, and a spray of pieces. List the cues that carry the property you want, in order of importance, and build those. His teaching example is “squishy,” because the word already implies a look, a sound, and a touch. The thing deforms on contact, especially with speed, and the movement has a sound. A full fluid simulation is more than the prototype needs. A simple squash and stretch, plus that sound, can carry the idea. He is frank that polish eats time. Spend that time on the cues that sell the physical impression, and treat a little shake or a short spray as enough to change a feel.

Metaphor changes the feel even when the tuning stays. His swap is a simulation-style racing game, Gran Turismo or Project Gotham or whichever one you know, with the car replaced by a huge balding runner spraying sweat. The structure, the tuning, and the function are untouched. The feel is different, because the player arrived with a model of how a car turns. People do this with animals too. His example is horse riding: a player says it does not feel like a horse, and then says they have never ridden one. The model was already there. For a prototype, iconic is fine, and a car still has to read as a car. Everyday objects set an expectation, and they are not the only subjects.

Rules are the last layer. Longer goals give the sensation of control a further meaning: from here to there, up the hill, collect a set of things, land a trick in a place. He wants those goals to be throwaway tests of whether the motion has any depth, meaning a sustained interaction that is more than the pleasure of steering in the simplest spacing. He is explicit that this starts to be game design proper, and that the goals should not be mistaken for a prototype of the whole structure. There is also a difference, in his account, between a goal you invent while noodling and a goal the computer enforces. Completing the coded one means meeting an impartial condition, and it implies some reward, however small. If you cannot find several goals that are enjoyably challenging, he treats that as a sign to change the mechanic or leave it. His closing line is that the feel is the first, last, and most common thing the player meets.

Juice is one of the pieces

Pichlmair separates juice from the whole sensation. Juice, and juicing, is abundant feedback on an interactive element. It is usually filed under polish, the surface treatment. It is one aspect of feel. The feeling of a game is not captured by juiciness alone.

Two talks carry the practitioner case for that feedback. Martin Jonasson and Petri Purho, “Juice It or Lose It” (2012), and Jan Willem Nijman, “The Art of Screenshake” (2013). Both argue that abundant feedback strengthens the connection in each moment, clarifies what is happening, and supports the player. Juice is easy to hear as cosmetic. The talks treat it as part of the gameplay. Pichlmair adds the limit. Making every moment feel good is not the whole of design. A game can want a varied feel, including an intentionally bad moment, because the point of that moment is the badness. Feel also changes from player to player and from situation to situation.

Lars Doucet, in the entry’s account of a 2016 piece, warns that abundant feedback can confuse, and that adequate juice supports the player. Douglas Wilson, the same year, splits the words. Capital-G Game Feel is the positive sensation of control Swink is after. Lower-case game feel is the feeling a game, or a piece of it, communicates in general. The lower-case term can cover unpleasant sensation, and it can cover games Swink’s avatar definition leaves aside.

Lisa Brown’s line, quoted in the encyclopedia from 2016, is the instruction that keeps juice attached to a decision. You are picking a feeling the game should communicate, and juicing that feeling.

Tuning, juicing, streamlining

Pichlmair and Mads Johansen’s survey, “Designing Game Feel” (2021), read more than 200 sources and sorted the work by purpose. The encyclopedia summarizes three. Tuning shapes the physical simulation: parameters for how objects behave, including movement, gravity, collision shape, and, in their grouping, screen shake. Juicing intensifies and clarifies: particles, freeze frames, bullet time, recoil. Streamlining makes the game act on the player’s intent by reading the input in context. Aim assist, coyote time, and button caching are the examples. A few techniques sit in more than one purpose. Bullet time, in a 2010 Rockstar San Diego game, slows the world while the player acts, which makes the action easier to read, exaggerates the feat, and can give a collision more heft, the way a Criterion Games crash in 2008 shows more of the physics in slow motion. The survey is a map of the research. It is not a quota of effects to add.

Hicks and coauthors, in a 2018 study the encyclopedia cites, interviewed 17 developers and, in that paper, treat juiciness as good game feel. Three components come out of the interviews: coherence of theme, gameplay, and feedback; the way state changes are exaggerated, focused, and given ambient response; and direct feedback that confirms the action across more than one channel without becoming ambiguous. The encyclopedia’s reading is that most of what those developers tied to good feel is the quality, fitness, and clarity of feedback.

The entry also points at craft write-ups that use the term for one feature: combat technique, the axe in the 2018 God of War, sound, genre practice, platformer parameters. Those are uses. They are not a checklist for a first game.

A footnote records Chris McEntee, designer on Rayman Origins, on the 3Cs: camera, character, and controls, in relation, as a core of feel design. Celia Hodent’s 2020 account, as the entry describes it, pairs that trio with Norman’s emotional design and sets game feel beside flow in the user experience. Swink’s 2006 essay says that, inside that paper, he treats Csikszentmihalyi’s flow and the industry word immersion as interchangeable. The encyclopedia’s note is that flow was built for activities other than video games. A prototype can aim at an absorbing moment without claiming the research term.

Rogers (2010) calls a varied set of feels friction, and Anthropy and Clark (2014) call the designer’s side resistance. Feel, difficulty, and access meet there. The MDA framework, from Hunicke, LeBlanc, and Zubek in 2004, puts feel in the aesthetic that dynamics produce. Keogh, Yang, Ehrndal, and Larsen arrive through embodiment, politics, and aesthetic theory, and the entry treats those approaches as tools for the intended feeling of an interaction.

What “good” was asked to include

The book, as the encyclopedia quotes it, says a game feels good when it has seven properties.

Predictable results: the action gets the response the player expects. Instantaneous response: the answer feels immediate. Easy but deep: minutes to learn, and a long time to master. Novelty: the result stays predictable, and the controls stay expressive enough to feel fresh across hours. Appealing response: the sensation of control is pleasing on its own, apart from the context. Organic motion: steering the avatar makes appealing arcs. Harmony: every element supports one cohesive sense of a physical reality that belongs to this game.

Pichlmair’s comment is that the weight sits on feedback and response. Some items need translating. A game without an avatar can still have a motion that reads as organic, and a game with very little graphics still feels like something. Only “easy but deep” on that list ties feel directly to rules, which is why the entry also treats learnability as part of access.

An earlier essay, “Principles of Virtual Sensation” (November 14, 2006), asks the same question and reaches for the same physical words players use: floaty, twitchy, smooth, slow, loose. The sections develop predictable results, novelty, traction in the first minutes, a low floor with a high ceiling, context, impact, and an appealing reaction even in an empty test. Predictable results is the cornerstone there. A mapping that flips at random leaves nothing to feel. Traction is the moment after the first success, when the player can see the challenge and judge whether their skill is a match. The two lists are his attempts to say what a good answer is. A first game can use them as questions. Did the verb do what they expected, in the same moment? Is the tenth try different from the first? Would you keep touching the motion with the level deleted?

The one verb you already have

The loop I use is one verb. Shoot, place, dodge, answer. Feel is the sensation of that verb, not a second verb added because the first one was dull. If the prototype needs two verbs before it is fun, the earlier page already treats that as a menu in disguise. Juice will not repair the disguise. Brown’s line still holds: pick the feeling that verb should communicate, and spend the feedback on that feeling.

The order from Swink’s garden fits the order you already have. The scheme names the input and the device. The response is one behavior: this button, this result, and what happens if the other action is already held. The space is one object the verb can meet, spaced against the motion, not a level. Polish is the smallest cue that sells the meeting, a sound or a deformation or a short shake, after the meeting exists. Metaphor is whether the object reads as the thing the player already has a model for. Rules, the start and the loss and the quit, come after the steering is a pleasure. A goal you code too early freezes a sensation you have not felt yet.

A game whose verb is “answer,” a puzzle or a text choice, still has a feel under Pichlmair’s wider term. Swink’s simulated space may be absent. The response is not. The player needs to know, in the same moment, whether the answer landed, and the feedback has to be the feedback for that landing. A shake copied from a platformer sells a collision this game does not have. Harmony, in the book’s list, is one perception. A cue from a different game splits it.

Intentionally flat or unpleasant feel is a decision, and it belongs in the sentence you are testing. If the moment is supposed to feel heavy, late, or harsh, write that down before someone plays, the way a playtest expectation is written so a stranger can fail it. Wilson’s lower-case term is the one that covers that moment. Swink’s capital sensation of control is the one that covers the verb you want people to repeat. Mixing the two up is how a harsh moment gets “fixed” with particles, and how a dull verb gets described as a bold negative feeling.

Watch someone who did not tune it. Swink’s picture, from Rad Racer, is a person leaning in the chair to push the car a little farther. Look for that lean, for a glance at the keys, for a retry that changes, and for a sentence about what they expected. Your own pass already contains the timing. If they cannot tell a hit from a miss without you, add feedback before you add content. If they can tell, and they do not want the second try, the verb is the problem. The screen can shake later.

Related: Game development, Unity vs Unreal Engine, Tingbase, Mini Games Arena, projects.

Common questions

What is game feel?

How it feels to control something in a game: how fast it responds, how it moves, and the feedback you get from sound, animation, and effects.

What is juice in game design?

Extra feedback on an action, like particles, screen shake, and sound effects. Juice makes good controls feel great, but it can't fix controls that feel bad.

How do I test game feel?

Watch someone who didn't build the game play it. Notice if they lean in, retry eagerly, or seem confused about what happened.