Sakurai on Creating Games: Specifications (18)

Masahiro Sakurai, the creator of the Kirby and Super Smash Bros. series, shares a wealth of game-development know-how on his YouTube channel “Masahiro Sakurai on Creating Games”. This article summarizes and restructures the content of that channel by topic.

That said, a summary is only an entry point. So much of the value — Sakurai’s own words, his real-world examples, his pacing, and the footage itself — can only be gotten from the videos. So please don’t stop at reading the article; I strongly encourage you to also watch the original videos embedded under each topic.

This category covers the specs that create play feel, such as hit-stop and jump mechanisms. Here we bring together the key points of both videos in the “Specifications” category (18 individual topics), structured in 2 parts following the 2 source videos. Each part opens with its explanatory video, so please watch along.

Part 1: Hit-stop, jumps, and designing feel (#01–#08)

This category deals with how to think about the mechanisms built into a game. More than planning and game design, it centers on concrete talk close to play feel, dealing with the computer.

https://www.youtube.com/watch?v=Edmd2xs9rZ4

1. Stop at the important moments!

Hit-stop is the staging where, the instant you strike an opponent, things freeze for just a little. With it versus without it, the feel is completely different. A very important spec.

What Sakurai calls “boss-stop” — freezing the moment a boss is defeated — was already implemented in the Famicom’s Kirby’s Adventure. The instant the boss falls, the screen flashes while freezing. It’s very important for getting sufficient feel.

Because it’s this important, you can attach it all over, regardless of genre. For example, scrolling shooters tend to make you go “huh, did that just hit me?” In 1986’s Fantasy Zone, the screen freezes for an instant when you’re hit by a bullet, heightening the sense of “oops.” This freeze staging, in fact, can be traced back to 1980’s Defender. Freeze firmly at the crucial moment to stage the feel of defeating, or being defeated.

2. Make the physical button feel weighty

Have you ever felt weight in a button while playing? Not the physical force feedback of the PS5’s DualSense, but the idea that the same physical button gives a different impression depending on what it moves on screen.

Control with the Famicom’s D-pad differs greatly in touch by the character’s movement spec. Not just whether motion is slow or fast, but inertia, jump spec, frame rate — many things influence it. The same button, yet this much difference in impression. This is a weapon of video games, he thinks.

A video game is played by receiving info from the computer with eyes and ears, and giving info with the hands. Carefully examining the “touch” you get from the hands changes how you assemble specs. In Smash, each fighter’s pace of movement and the exaggeration of an attack’s wind-up are examined not just for mind games but for feel. In fact, since the analog-pad spec differs by platform, the so-called “dead zone” (the minimum value to recognize tilt) is set again every time.

3. The jump mechanism

A character jumps. It seems simple, but there’s a lot to think about. Generally, while in the air, a constant fall acceleration is applied continuously. Give an initial upward velocity (say 5), add fall acceleration (say 1) each frame, and the velocity drops 5→4→3…, reaching 0 at the peak, after which it accelerates downward and falls.

This jump spec should be varied by the game’s purpose. Super Mario rises slowly, making it easy to vary the height by how long you press (perhaps because it’s a game where the height to shoot a fireball matters). Ghosts ‘n Goblins allows no mid-air horizontal movement; in exchange for the downside of a fixed landing point, it generates mind games of thinking jumps through. Smash’s jump has a “crouch” go in just before, and you can’t make it 0, because removing it would jump before an up-smash. Note that Smash Ultimate adopts a hidden touch: give an extremely strong initial rise speed, a strong fall speed for just a few frames, then return to normal. It can cope even if the player’s reaction is a bit late, and gives good rhythm — a fine spec, he says.

4. Eight hit-stop specs

Hit-stop, where both sides freeze for an instant when an attack lands. As a memorable example of putting it in explicitly, Sakurai cites Final Fight. Smash’s hit-stop has many special specs not found in other games. Saying “this is a trade secret, but…” he reveals the eight.

For example, ⑤ scales hit-stop larger the higher the attack power while applying a coefficient, tuning Marth’s smash attack so that landing the tip feels superb, but the base has no feel. ⑦ is the ultimate special spec: when a sword-wielding fighter slashes, the blade’s tip moves during hit-stop at a tiny velocity of less than one frame. Each such accumulation is what reconciles feel and smoothness.

5. Don’t glue the focus point

A 3D game’s camera has a “focus point.” Usually invisible, but it always exists in any game with a camera. You can look all around centered on a character because the focus point is roughly at the character’s position. But Sakurai says you mustn’t glue the focus point to the character.

In Super Mario 64, while walking you’re at screen center, but when you jump, you leave the center and your relative position goes up. This lets you tell how high you jumped. It seems obvious, but it isn’t. In Armored Core, up/down/left/right controls shift the AC slightly off center, making dodged bullets easier to see. Kirby Air Ride, too, shifts the machine a bit off the focus point during a drift, heightening the sense of control. Rather than just following, give it staging that suits the game.

6. Slowness is a sin

Games have all kinds of personality, and tastes vary. So you can’t flatly deny something just because you don’t find it fun. But there’s nearly one thing no one can love. That’s “being slow.”

Not just plain movement speed, but everything taking time, long stretches of doing nothing — in other words, how much time you spend not playing the game. Making players wait when there’s something they want to do or that’s fun is something you should consider very bad. There’s time you can’t shorten, like loading, but time you can shorten should be cut down to provide denser time. “Palate cleansers” like cutscenes are better to have, but wasted blank time should be trimmed. When creating, Sakurai is conscious of the real time spent operating the game, even feeling “sorry” for time when he’s entertaining no one.

7. The unit of speed

Speed governs all sorts of things in a game — a very important element. Speed = distance ÷ time, and knowing two of the three gives the remaining one.

In the sprite-based era like the Famicom, speed was expressed in DPF (dots per frame) — how many dots you advance per frame. Mario’s top speed is probably around 1.x DPF. But since the Game Boy, Famicom, and Super Famicom have different dot counts (size of the ground), the same setting yielded different speeds. In 3D games, defining characters and fields in real units lets you use realistic speed units like km/h directly. Kirby Air Ride uses km display, but since Kirby is assumed to be about 20cm, even a low km/h is actually quite fast. Without being able to show speed in the spec, you can do nothing, so unifying the sense of distance and unit of time in a game is essential. Cultivating a sense of speed in games matters.

8. Behavior at cliff edges

Many games let you fall off cliffs, and that spec varies too. The fall unconditionally spec was common in old games, but not uncommonly it’s bad. The fall while moving, don’t fall during an action spec (falling on every attack step feels like an action cancel, which is unpleasant), and further the fall/don’t-fall decided per action spec (forward dodge doesn’t fall, a charge attack does) are common.

In Smash, Fox’s Illusion stops at a cliff, and Falcon Kick flies off the cliff (but doesn’t fall near its end). Without watching speed carryover, you can suddenly kick off at high speed. In 3D games, specs like fall when you push hard against the cliff, and interpolate so you don’t fall when the cliff edge ahead is shallow are also effective. Cliff edges, where you’d assume falling is the norm, allow for this many considerations. Don’t miss the discomfort during test play; aim for the appropriate spec.

Part 2: Randomness, handicaps, average vs. mediocre, online updates (#09–#18)

This part centers on specs that sway a game’s fairness, fun, and staging — randomness, handicaps, balance tuning, online updates.

https://www.youtube.com/watch?v=tqF53jjKdq0

9. The map is a game screen too

Playing various games, you sometimes feel “I’m basically playing on the map screen.” Walking in 3D space, yet only watching the corner map. It’s a waste to play in the corner of a vast, beautiful screen.

What stays with Sakurai is the Mystery Dungeon series’ display that overlays the map semi-transparently across the whole screen. Press a button and the game screen vanishes, viewable as map only. Beyond enriching the map, increasing the main screen’s information is also effective. Like a racing game’s guide display showing the degree of a corner without watching the map, or making turns clear via terrain depiction. It matters to account for the player’s awareness of where on the screen they’re watching while playing.

10. An unexpected finished form

Sakurai, swallowing his shame, reveals experiences where something unanticipated occurred in the retail version, leading to an unsatisfying result.

Meteos sped up due to late-development processing optimization, changing the difficulty curve so it became harder than intended. Mushiking had processing speed differ between dev hardware and retail, so hunger came faster. Smash Ultimate, in the released version, somehow had the CP feel stronger; multiple staff agreed, but the cause stayed unknown. As long as you’re dealing with computers, you don’t know what will happen. Other developments going well, too, may be a tightrope walk of precarious balance, Sakurai says (Ultimate could be handled via online update, but in an era without it, nothing could be done).

11. The flick input

When the Nintendo 64 first handled an analog stick, what differed from the digital D-pad was that “direction and degree can be taken steplessly.” Perfect for walking in 3D space. But Sakurai focused on one more thing: you can take the time for input to reach the edge. This is what he dubbed the “flick input.”

Until then, dashing needed two-step operations like tapping the same direction twice or combining a button. Varying the strength by how fast you tilt the analog stick is emotive and easy to understand. Smash’s tangible controls also come from this. Smash Ultimate even lets you tune the flick-input speed. In Kid Icarus: Uprising, the dash doubles as a dodge, meshing with the risk and reward of drawing in the opponent’s attack, and made controls simpler. Don’t adopt new devices blindly; it was good to create unique controls with a general-purpose analog stick.

12. How to Play

Mastering a complex game’s controls from scratch is extremely hard. So a tutorial is essential. But you don’t want a tedious tutorial. The worst is making people read long text like a manual. Next to avoid is operating on an inorganic practice field (you’ll be thought “just let me play already”).

Fairly common is placing text along the first stage’s path to read and then having them practice. Smash only attaches a demo movie, which isn’t by preference but the result of repeatedly failing to implement plans to have you learn the knack through controls. The original Smash’s “Break the Targets” was also a placement aimed at mastery. The point is to actively let players play while learning. Note that Sakurai still regrets, as “the most frustrating,” that a mode to learn techniques while playing, planned for Smash Ultimate, was dropped for schedule reasons.

13. Randomness adds color

Randomness (a mechanism that decides elements haphazardly), used well, adds color to a game. The good point is that it makes repetitive play less boring and keeps freshness with a different development each time. It seems to lose strategy, but it actually grants the strategy of “coping with various situations.”

The bad point is that it isn’t fair. When luck changes the outcome, the effort you put in is ruined and unconvincing. There are RPGs that parameterize luck (long ago, 1981’s Wizardry), but the priority is low. Smash has recently trended toward removing random elements from fighters (Peach’s smash became choosable). However, Hero’s “random command menu” is OK because whether you use it well depends on skill. Leave varied randomness in items to stir up the field and leave a chance for anyone to come back. For a “laugh it off and move on” game, this direction fits. Note that since computers can’t make strict random numbers, referencing something indeterminate like the built-in clock is good.

14. Handicap boost

In versus games, an evenly-matched development is the most fun. But skill gaps are cruelly large. So versus games sometimes add handicaps. The extreme is Mario Kart. The lower your rank, the more the speed-up boost, plus item favoritism. It tends to drag down the leader and can’t be called fair, but for a casual game, you should definitely put it in. The first-place player staying first forever isn’t fun.

Smash, however, basically doesn’t pander. While leaning casual, its policy firmly reflects skill, and there’s no rule like “lose and your attack power rises” in principle. As common systems, the leading player can’t easily grab the Smash Ball, and the Final Smash meter fills faster for the one losing — about that much. The “anything goes” rule that creates outcome variance via terrain, items, and stages suits party games. And per Sakurai, the “human handicap” of deliberately taking hits for kids or beginners is the greatest effect of all. It’s in a home setting that you can iron out the bumps.

15. Average and mediocre are the same

In specs and parameter tuning alike, Sakurai wants you to hold the idea that “average and mediocre are the same.”

In Smash’s balance tuning, looking only at strength, weaknesses get patched and strengths shaved, so performance keeps averaging out. That’s the least fun of all. So you supervise in a direction that doesn’t kill identity, like enlarging weaknesses to match the strengths. Ultimately, making it Mario vs. Mario gives perfect balance, but that’s no fun. Clashing different things generates various mind games.

Here Sakurai reveals data he calls the world’s fairest: the win rates of Smash Ultimate’s online matches.

The 87 fighters fit within a narrow range, from a high of 51.43% to a low of 47.118%, in a beautiful gradation. Data you could call “even the lowest-ranked fighter has a chance.” With no weaknesses and no strengths, it doesn’t feel appealing. The message: those doing balance tuning should add sharp contrast so characters have identity.

16. Online updates

About online updates, Sakurai candidly says “I’m grateful.” This time, four points are raised from the maker’s side.

First, recent games are too complex. A complexity said to be 100x or 1,000x of old, where huge data generates un-digital-like fluctuations. Just having everything work correctly is miraculous. Second, you can’t play the final version during development. Touch it up and it’s no longer the final version, so the time spent with an unmodified final build is very short. Third, the market’s testing power is on another level. A 1-in-1,000 bug is hard to find during debugging, but out in the market the number of thought-cycles is on a different scale — and it gets uploaded to video.

Fourth, balance tuning is ultimately a voluntary service by the maker. Fatal bugs should be fixed, but tuning binds the director and others and costs money to do. Even so, making things better shouldn’t be criticized, he thinks. But also: if a player quits at the start, that’s the end. Better to be in a better state from the very beginning.

17. Make screen transitions swift

Screen transitions should be done as snappily as possible. As discussed elsewhere, slowness is a sin in games. A sleepy, slow fade-out, a long blackout, a long time before returning control — all of these alone rob the player’s precious time. A bit impatient is just right (though with no staging at all you get confused, so “swift” is the basis).

That said, as a maker’s circumstance, many works use the blackout for load time. The first Resident Evil’s door staging doubled as loading while also being a horror effect — the door opened a bit slower in rooms with zombies, reportedly. The first The Legend of Zelda’s screen-edge scroll also had the merit of clarifying positional relationships. Give it a purpose suited to the game while not spending the player’s time wastefully — that’s the basis. If the tempo fits, mastering transition staging like wipes is one hand too.

18. What happens if you offset the center of gravity?

Adding emotion to mechanical “straight movement” or “rotation at the center” with just a little ingenuity. This too is an important spec.

The cannonball that appears in Kirby wasn’t interesting as expression when simply flown straight, so they shifted the sprite’s display position to make one revolution over 6 frames (internal name “wobble shot”). Its presence became far stronger, and it paired well with the Game Boy’s afterimage-leaving screen.

When throwing a bomb in Smash, the rotation axis is shifted a bit from the bomb’s center to give an unstable feel. Fixed to the central axis, it would have looked more inorganic. Pyra’s side special “Blazing End,” too, at first had the sword’s grip as the rotation axis, but shifting it slightly gave a difference in expression. Animation is “putting a soul into things that don’t move.” When you make it as a spec, straight lines and constant-speed rotation tend to become the default, so you’d want to apply ingenuity so it doesn’t go bland by mechanical common sense.

Summary

We’ve now gone through both videos and 18 topics of the “Specifications” category. Finally, let’s recap the key points of each part (= each source summary video) by individual topic.

Part 1: Hit-stop, jumps, and designing feel (#01–#08)

Part 2: Randomness, handicaps, average vs. mediocre, online updates (#09–#18)

There’s a lot here, but I hope you’ll revisit it starting from whichever part caught your interest. Please also check out the related categories.