Boghog's bullet hell shmup 101
This is the wiki version of this doc
This doc covers the fundamental aspects of designing a bullet hell (danmaku) shmup. This is heavily skewed towards CAVE’s style of games, but a lot of the things discussed here can be carried over to other styles.
Knowing these fundamentals is important even if you’re going to break every rule in the book, because it helps you make informed choices instead of taking shots in the dark.
I’ll include examples when appropriate but for practical purposes, I will avoid too many counterexamples, edge cases and other complications. The goal is to give a good idea of how to make a typical CAVE-style shmup. If you want to branch out past that & make something different you’re on your own!
And don't forget, first hand experience is everything! Play shmups a lot, you don't have to get 1cc's or good scores but getting to a point where you can watch a good replay of your favourite game and understand what the players are doing is a must. Game design requires good intuitions for what feels right, and the best way to build that intuition is through experience.
- 1 MOVEMENT & SPACE
- 2 SHOOTING
- 3 LIVES, BOMBS AND RECOVERY
- 4 BULLET PATTERNS
- 5 ENEMIES
- 6 LEVEL DESIGN
- 7 SCORING SYSTEMS
MOVEMENT & SPACE
The Play Area
This is where all the action happens. The play area can either be contained by the screen, or it can be wider thanks to horizontal panning seen in many vertical shmups. The size of the play area is relative to the size of the game’s objects and hitboxes.
The tiny hitboxes of Danmaku games make for relatively large play areas which allows them to fill the screen with bullets. This has downsides however - large play areas funnel games into relying on lots of projectiles for challenge. It makes it much harder to rely on simple patterns.
A game's movement speed and shot type has to be tuned to match the size of the play area, narrow areas can afford to (and can benefit from) slower move speeds and narrow shot types to emphasise small differences in positioning and give enemies time to do their thing (examples : Gunbird 2, Dragon Blaze). Wider play areas tend to use faster speeds, wider shots or some kind of multi directional weapons to compensate for the large distances the player has to travel (examples : Mars Matrix, Under Defeat HD).
The most fundamental source of challenge in danmaku games is identifying, predicting and manipulating different bullet trajectories and making precise movements to dodge bullets and control screen space. Because of this, giving the player as much control, consistency and awareness as you can is the top priority, the movement should feel seamless.
For consistency’s sake, it’s best to avoid different directional movement speeds, such as the ship moving faster on the x axis than the y axis, or faster diagonal movement. Don’t forget to normalise your diagonal movement! There are some great games with faster diagonal movement like Battle Garegga and Armed Police Batrider, but they are rare.
Movement inertia should be avoided entirely because it adds an unnecessary layer of lag to the movement that players have to adjust to, without adding any interesting gameplay dynamics. Even the smallest amount of acceleration will be noticeable to experienced players. If your movement doesn’t “feel smooth”, then it’s most likely the result of lacklustre visuals rather than a lack of inertia. Some tricks that can make movement feel smoother - beef up your shot/rate, add an afterimage (see Symphony of the Night/Megaman ZX), match your ship’s banking animation speed to movement speed, have the ship leave trails (see Danmaku Unlimited 3), give your ship’s options (floating bits near your ship) some inertia.
Danmaku games often feature a focus shot mechanic, which was popularised by CAVE. The player’s ship has 2 states with their own speed, a fast wide shot mode (tapping the button), and a slower focus shot mode (holding down the button) usually represented by a laser or another kind of concentrated shot.
This is useful because it gives the player more control and creates some basic but rewarding gameplay dynamics - the players have to think whether they need to use fast movement to quickly get into position, use slow movement to make dodging more precise, or accept slow movement for the sake of additional damage-per-second.
The transition between normal & focused speed is usually (but not always, see Touhou) interpolated to create a smoother, more gradual transition. The speed difference between the normal and focus shot itself varies depending on game & ship (focus being ⅔ of the original speed is a good middle ground).
The gameplay effects of different speeds are highly varied, the usual archetypes you have are fast, powerful ships with narrow shots, or slower weaker ships with wide shots.
Hitbox and visibility
During gameplay, players won’t have the time to look at their ships, instead focusing on enemies, their patterns and places they are moving to. They will roughly estimate the ship’s position based on the stream of bullets they shoot out, the general silhouette of the ship & additional visual elements (like HUD elements, flashing colours) near the ship. As a result, giving players thick, fast, noticeable bullet streams can not only enhance your game’s feel, but also the visibility. Perfectly centering the hitbox is also important because it keeps things consistent.
Shmups are all about shooting. Above all it’s important to make the act of shooting feel satisfying. Achieving a decent game feel isn’t too difficult, but there are a lot of little things to take into consideration. A general bit of advice - when you’re making a game’s objects move, you become the game’s animator. The only difference is that you animate things through code rather than by redrawing frames or moving things around. As a result, studying fundamentals of animation & practising will pay off immensely.
On the mechanical side, the key principle for achieving a good game feel is to skew things in the player’s favour in subtle ways. Give the player a hand & compensate for their small positioning/timing mistakes, while focusing on punishing big ones.
Polish Effects & Small Mechanics
The first thing you want is speed.
- Speed is good for conveying force and making impact feel stronger. A fast bullet is a powerful bullet, especially if you combine it with an appropriate bullet splash effect.
- Speed helps create immediate feedback which is important for enhancing the player’s sense of agency. When the player presses a button they expect results.
- Speed helps the player keep track of the ship’s position. The faster you can update the bullet stream to reflect the player’s current position, the better they can estimate it.
In animation, a good way to convey the feeling of speed is by using motion blur (smears, trails). Length will create the illusion of motion and make bullets feel even faster. This can be taken to ridiculous extremes and still look good. The opposite will likely read poorly - short sprites will clash with fast travel speed and create a disconnect. As a general rule of thumb, you should always consider the projectiles’ travel speed when deciding the length of its sprite.
The next element you want is density. Make big, fat projectiles, huge messy streams, cluster bullets together and don’t concern yourself with making things too neat and organised. All of this makes the player ship feel like a force to be reckoned with. Players want to feel powerful, they want to feel like they’re wiping out everything in their path and overwhelming the enemies, even as the game is kicking their ass. Dense bullet streams create this illusion.
Dense patterns help the players estimate their position, smoothen out imperfections and let the player’s mind imagine more interesting bullet streams than what’s actually on the screen.
Giving the player’s shots huge hitboxes, and giving enemies huge hurtboxes makes the game feel better by compensating for the player’s minor positioning/aiming mistakes. The player’s shots shouldn’t have massive gaps or dead zones, and their emitters should be low.
The on screen shot limit is when games only allows a limited amount of player projectiles to be on the screen at once, it refuses to create new ones until the old ones exit the screen, hit an enemy or are otherwise destroyed.
Born out of hardware limitations, the mechanic has become a staple of the genre because it perfectly meshes with the player’s movement and creates some very fun natural gameplay dynamics. The closer the player is to an enemy, the faster their shot rate (and DPS) will be.
The prominence of the shot rate is different in every game. Games/ships with high fire rates and low on screen shot counts encourage a very aggressive in-your-face kind of playstyle while the opposite creates more dodging oriented games.
It should be noted that in addition to this, some games have other forms of proximity damage, such as Dodonpachi’s ship aura, which surrounds the ship and does tick damage. Other unique ways of creating proximity damage are worth experimenting with.
Power ups are all about trickery and illusion. Increasing the damage in a straightforward manner will most likely lead to balancing issues. The player's level 10 shot cannot be 10 times more powerful than their default shot without something breaking.
Shmups get around this problem by simply lying to the player about their power level, increasing damage by a very small amount (for example x1.1). This practice is normal and is seen in most shmups from Toaplan to CAVE.
Numbers aside there are other ways to make shots feel more powerful. Increase projectile width and height, increase their speed, make them look more saturated, add details to the shots, make damage sounds more powerful. You can even decrease the damage values but add additional projectile emitters, like Gradius-style trailing options. Anything that makes the shot more satisfying without causing balance issues or making the lower levels feel terrible is fair game.
LIVES, BOMBS AND RECOVERY
Shmups tend to be pretty strict with how they handle lives - in CAVE games you tend to get a starting stock, a couple of extends and a 1-UP item. This strict approach has its benefits as it forces the game design to be very tight since any mistake will be severely punished. The games are, with some exceptions (true last bosses) designed to let you avoid every single bit of damage with some practice.
Shmups with robust, large health bars tend to lack this rigorous design, which is why health bars are a bit of a taboo. That does not mean that they’re bad, however - multiple CAVE games like Guwange, Deathsmiles and Akai Katana use them quite well.
Because each mistake is very costly, the games give you a breather if you do die. After each death, all enemy bullets are cancelled to clear up the screen. After a short while, this cancelling stops but the player is still given a couple-few seconds of invincibility to kill some enemies & reposition themselves. Not destroying bullets helps with the repositioning process since players will know what to expect. Generous invincibility is important to prevent chain-deaths.
Bombs are a multi-purpose resource, they can be used defensively (called panic bombing) and offensively to safely & quickly kill bosses. It’s beneficial to add a small buffer so that if the player bombs within a couple or a few frames of their death, they can nullify their death. Dying on the frame you bombed always feels frustrating.
The overlapping projectiles and patterns, enemy waves, various particle effects and items make for very chaotic and busy screens. Yet despite this density of objects and information, danmaku games must have exceptionally good visibility to guarantee a fair, non-frustrating player experience.
The world of art provides a very helpful concept when dealing with visibility - VALUE. Value refers to the lightness/darkness of any given colour, it takes hue/saturation/brightness into consideration, as each part of the hue is unique. It can be grouped into 3 general categories - highlights, midtones and shadows.
While looking at the values of different bullet sprites, you may notice a pattern - they put light & dark values side-by-side. The bullets often have very bright elements (the glowing cores) right next to dark elements (borders, sometimes inner circles/lines). This is how you maximise visibility by using values.
Low contrast backgrounds that rely primarily on midtones also help with visibility since they let you freely use extreme values for important elements.
Chunking patterns is vital for visibility. Players try to predict bullet trajectories and move accordingly, and chunking helps to telegraph this. Try to group bullets up into lines and other clear patterns, single stray bullets are hard to read and can often feel unfair. Bullets with unusual, hard to predict trajectories may need extra effects like trails to help players out.
Animation also helps bullets stand out. Looking at CAVE bullet sprites will quickly reveal all kinds of wobble and ripple animation which catch the player's eye and give each bullet a unique identity.
Last but not least you have depth sorting. Enemy bullets should always be drawn on top of other game objects such as player sprites, projectiles, items and explosions. Use bullet size and speed to inform their depth. Smaller, faster bullets should be drawn over bigger, slower bullets. Single bullets or small chunks should be drawn over bigger easier to read chunks.
When you break things down, danmaku games only have 3 simple types of patterns, they are as follows :
- Aimed. The trajectory of the pattern is based on the player's position on screen. Good for pressure, allows conscious manipulation by the player. Can be quite dynamic due to inconsistency in the player's inputs.
- Static. The bullet trajectories are predefined and do not change based on the player's position. Good for creating obstacles. Static patterns give the designer a lot of control, letting them flex and make beautiful and cool patterns.
- Random. The bullet trajectory is randomised. Keeps things fresh. Has to be used carefully because it can create unfair situations.
The x/y coordinates of the emitters can either be preset and unchanging (which guarantees that the patterns are clean and consistent), or they can change based on the enemy's movements (which allows you to create very dynamic and unpredictable patterns by bending and distorting the patterns).
Controlling screen space makes up the majority of shmup gameplay, which means active manipulation of bullets and enemies by the player. Keeping this in mind, you can identify some common enemy roles:
- Pressure. The shots of these enemies put constant pressure on the player and force them to move around.
- Area denial. The shots of these enemies block off parts of the screen and make movement more difficult.
- Direct challenge. These enemies are usually elites that challenge the player directly with dense, difficult patterns.
Levels mix and match these enemy types to create interesting challenges.
The player's engagement with enemies should be dynamic. You should never let the player feel like their actions don’t matter. A good, dynamic enemy design would be one which plays out very differently based on how well the player’s doing. Killing enemies quickly should be hugely beneficial, and leaving them alive should be rather dangerous. The last thing you want is enemies which are hard to kill, but also not very dangerous if left alive.
Think of enemy design in terms of roles given the player’s performance.
- Optimal Kill/Intro. What happens if the player kills the enemy immediately. Do you want to make sure the enemy shoots at least a little bit no matter what, or do you want to allow the player to instantly take them out?
- Average Kill. The enemy’s normal, intended behaviour.
- Slow Kill/Safety Net. What happens if the player is slow in killing the enemy. Do they linger on the screen and trap the player, or will you add some safety nets and have them fly away? Perhaps they could change their pattern to be less aggressive but still problematic.
An enemy’s priority has to be taken into consideration as well. Whether an enemy is high or low priority is an emergent property that depends on too many factors to break down into simple rules, but here are some things that tend to work:
- High HP. Bulkier enemies tend to command attention because they block player shots and can’t be easily cleared out whenever the player wants.
- Dense patterns. Patterns which are difficult to dodge and block off parts of the screen will make enemies exceptionally dangerous especially when there is a high amount of overlap.
- Wide cone. Enemies that have aimed shots that spread out over a wide cone are a lot harder to control, and as such more dangerous.
High rate of fire. This makes enemies particularly nasty and prone to trapping the player, encouraging players to dispatch them ASAP.
Some rules of thumb:
- Enemies, especially popcorn enemies, should not have much more HP than is needed to fulfil their function.
- Safely approaching enemies for a kill shouldn’t be disproportionately dangerous, it should never be more beneficial to ignore enemies and let them leave the screen than efficiently killing them.
- Enemies should have big hitboxes and slower predictable movement. Dense and difficult bullet patterns make up most of the challenges in danmaku games, so making enemies needlessly elusive and difficult to hit will likely cause frustration for the player.
To keep enemy health low while guaranteeing they will shoot at least once, it may be desirable to cheat and only make them vulnerable after that first initial shot.
Polish Effects & Minor Mechanics
A good enemy is fun to hit and kill. Some ways this can be achieved are:
- Hit indicators. Enemies should react to getting hit or else they will feel like ethereal forces rather than objects that exist in the game’s world. This can take the form of simple flashing, damage particles/effects or shaking.
- Subtle hit sound effects. The player needs to feel it when they’re doing damage to enemies.
- Big, meaty explosions. Explosion sprites should be large, significantly bigger than the enemy sprite. Different explosion sprites, patterns and extra debris are always welcome.
Some minor mechanics which help make games feel more polished:
- Bullet sealing. This is an anti-frustration mechanic that makes it so that weaker ground enemies don’t shoot if the player’s sitting on top of them/is very close to them.
- Dead zone at the top of the screen. This is a small space near the very top where enemies cannot be killed or damaged. This prevents the player from killing off-screen enemies they can’t even see.
- Ceasefire zone. The invisible part of the screen that renders enemies passive. This is primarily used for ground targets such as turrets and tanks, which drift to the bottom of the screen if left alive.
Danmaku level design might seem very simple on the surface, however there are many nuances to it that make it both difficult and interesting. At its core, it's just like any other type of level design - the goal is to build challenges that encourage the player to play the way you want them to, while leaving plenty of room for experimentation.
Flow should be the core goal of Danmaku level design. Flow is a feeling the players get when experiencing an uninterrupted sequence of smooth movement. It is achieved by guiding the player towards the intended route with the use of soft incentives & clear telegraphing and making the movement itself interesting. The player’s movement should be frequent. It should be smooth, one action should flow into the next; there needs to be a constant sense of forward momentum. It should be varied, testing the player in different ways - quick sweeping zig zag motions, precise tap-dodging, aggressive point blanking at the top of the screen and defensive dodging at the bottom. And it should take advantage of the screen real estate, clustering the action on one side of the screen is wasteful and unexciting.
Because danmaku games have full freedom of movement, it can be difficult to guide the player through the levels. You must rely on soft incentives, which requires a good understanding of how players move, what they prioritise and what they look at when playing danmaku games.
Here are some common player behaviours which you should keep in mind:
- Players will naturally stay closer to the centre of the screen on the Y axis. It gives them decent DPS and allows them to misdirect bullets more effectively while giving enough time to react to oncoming bullets.
- They will get close to high HP enemies to do extra damage whenever they get the opportunity. Usually when said enemy is preparing to shoot. Then they'll tend to go back down to dodge the volley of shots.
- As long as popcorns keep spawning in, the players will naturally move side to side as they keep streaming bullets.
- Dodging bullets at awkward diagonal angles or by mainly using up/down movement in one side of the screen is reserved for special occasions or last-resort scenarios, players will rarely do it unless they absolutely have to because it's harder to read bullets travelling diagonally, it requires more awkward inputs and it limits their offensive ability/dodging space.
The Toaplan Pattern
The Toaplan pattern is a very common enemy spawn pattern seen in most vertical shmups, but is used so frequently and so blatantly in Toaplan’s games that I came to associate it with them.
To understand the pattern it’s best to view the screen as a bunch of lanes. Their exact number isn’t too important but 5-7 tends to be the norm due to the sizes of enemies.
The idea is to spawn each enemy on the opposite side of the screen from the previous one to force the player to move & create a kind of sense of rhythm.
The Toaplan pattern works best with high priority enemies because they let you control the player’s movement and predict where they will be at any given time. It serves as the core “layer” of your level design that everything else is built around.
The Toaplan pattern creates a strong, reliable centre of gravity that pulls the player and makes them more predictable. To emphasise this and create a sense of rhythm, some amount of overlap is required. Enemies must spawn in relatively quick succession so that the player won’t be able to linger in one area for long. This also gives the enemies some time to breathe.
Overlap turns a series of disconnected enemy spawns into a cohesive level with its own sense of flow and interesting risk vs reward dynamics. The waves will affect each other by throwing off the player's positioning and blocking parts of the screen with projectiles, they give the player's actions more meaning as a result. If the player kills a wave quickly they will be rewarded with a safer, cleaner screen right before the next wave hits. If they fail then the wave overlap will force them to try and recover from their mistake. It gives the players good reasons to care about taking out all of those high HP enemies and popcorns as efficiently as possible.
You can augment your core layer by throwing in a bunch of weak popcorn enemies. Their most basic use is serving as little obstacles clustered on the path to the player's destination, which will most likely be enemies from the primary layer. These enemies will give the player extra things to do and worry about. The more the waves overlap the more chaotic the game can become as the shots of popcorns will block off parts of the screen and change how you approach future encounters and patterns.
Pacing, Variety and Extra Touches
Despite how short and dense danmaku games are, there is still a lot of room to make the pacing dynamic & memorable. Dynamism, smart repetition & variety are important because they smoothen out the learning process, making otherwise unmemorable encounters stick in the player’s mind.
Players like patterns and controlled repetition. Reusing the same enemy for several encounters back-to-back can make a part of a level more memorable, as long as there is sufficient variety in the encounters and how they challenge the player, and they don't overstay their welcome.
Constant intensity, heavy layering/overlap and high levels of challenge will blend together and wear out the player. Creating slight variations in the game's intensity levels by tweaking and emphasising those aspects of level design to different extents can help make the intense parts stand out even more.
You want some unique set pieces to punctuate levels and create landmarks the player can use when they are learning the layout. Big mid boss like enemies, background events and even unique props are very useful here.
Adding some interaction between the enemy layer and the background layer helps make your level feel like an actual physical space. This can be achieved with destructible scenery, hidden bonuses and various polish effects like ground explosions leaving behind craters.
Scoring systems are integral to shmups, the genre developed them more than any other. There are so many tightly made robust systems that it often feels like everything's been done already, and done well.
They add depth to games and make them more replayable by giving players more meaningful elements to learn & clear constantly shifting goals. They create dynamic difficulty by having players take on greater challenges as they get better even in earlier levels. They make the games more engaging on a moment to moment basis by giving players more to consider/think about/execute.
Scoring systems can add a layer of resource management to the games or give existing resource management more meaning. They connect different encounters into a cohesive whole and give players long term goals to work on beyond survival. And they can work as a way to gauge the player's skill growth over time.
Shmups have a very simple core layer that's transferable between games, scoring systems are what make the games truly come into their own.
Even if you, as a developer, have no interest in adding & fleshing out scoring systems, you must understand what they bring to the table and figure out good ways to recreate it by other means.
Constructing a Scoring System
When creating scoring systems, it’s good to think of them as sets of conflicting goals & priorities. This is what makes scoring systems dynamic & deep and forces players to make decisions and plan their routes.
The most basic example of this conflict is between scoring and survival - players want to stay safe to survive, but scoring forces them to go out of their way & do risky stuff to reach secondary objectives.
These conflicting goals also exist within scoring systems too. If your scoring is item-based it may be impossible to grab every single item the game can generate, forcing players to decide what to prioritise. If your game uses finite resources for scoring, then players have to figure out where to optimally use them. The fact that players can’t have everything forces them to approach the games more thoughtfully.
The principle is the same when creating any kind of game depth - create sets of conflicting short term & long term goals with some overlap and force players to figure out what to prioritise when.
Incentives & Tangibility
Scoring on its own can be a rather abstract mini-game which is detached from most in-game objects & natural player behaviours. As a result, scoring systems can feel arbitrary and meaningless to players. To counter this, games try to take advantage of intuitive incentives while making scoring elements more tangible & connected with the player's performance.
Incentives are natural/intuitive when they are either backed by very strong audio-visual feedback (making items tangible helps), when they take advantage of behaviours players are conditioned to, or behaviours that naturally follow from the rules of the game (such as killing enemies efficiently, or going fast in a racing game). The more in line with the feedback & rules of the game your incentives are, the less arbitrary they will feel. Going fast to increase your score in a racing game will feel natural, going in circles at the starting line will not.
The ideal scoring system is one that players fall into from playing due to incentives being perfectly lined up, one where players can gauge their performance without looking at numbers, and one that’s deep & challenging.
Linear vs Exponential Scoring
Linear vs Exponential refers to two types of scoring system, they should be viewed more as metaphors rather than taken literally.
Linear scoring systems are straightforward - you kill an enemy/pick up a scoring item, you get a set number of points. Typically these scoring systems have little to no punishment for missing an individual enemy/scoring item.
Once things like multipliers, big cash out moments & end stage bonuses are introduced, the scoring gain becomes exponential. Exponential scoring systems typically come with harsher punishments for failure. Your chains might break & your bonuses will reset/disappear.
- Reflect the player’s skill more holistically. They don’t disproportionately punish mistakes during specific sections, or make longer stages/specific parts disproportionately lucrative.
- Minimise frustration. Linear systems let you recover from mistakes with exceptional performance.
- Provide high risk, high reward. This can be very exciting and gives the games a strong “one more try” quality.
- Offer very clear feedback when the player is doing well/poorly. Sudden massive boosts of score are a good way to indicate that the player is doing something right.
Both styles have their own pros & cons. Games are rarely one or the other, they usually limit the exponential elements (such as capping multipliers) to create a more linear curve and add various small multipliers to further optimise fairly linear scoring systems.
Aggressive vs Defensive
Defensive Scoring Systems The game attacks you with its scoring. Whenever you perform an action like killing an enemy or picking up an item, the games give you some kind of bonus and start a timer forcing you to continue certain actions in fear of losing said bonus. Being passive is massively punished. Example: Dodonpachi
Aggressive Scoring Systems You attack the game. Extra score has to be actively earned by performing certain actions. The game does not pressure you into continuing combos, being passive is simply not rewarded. Example: Giga Wing
Games exist on a spectrum between these two styles. Certain games like Guwange have strong defensive & aggressive elements - your chain is timed, but it also doesn’t increase much unless you aggressively earn coins.
Depth vs Clarity
It may be tempting to go nuts with all kinds of multipliers, sub multipliers, varied base values, several levels of resource management & more in order to create a deep system that’ll keep players occupied for a long time.
This has massive downsides because as you’re increasing a game’s mechanical complexity in an effort to maximise depth, you’re making your system hard to understand and as a result preventing players from forming a solid game plan. Focusing solely on adding depth without increased complexity is a potential solution, but not a practically viable one.
To solve this problem, games create a hierarchy of systems/goals and limit depth by creating more clear discrete states. To do this mechanically, scoring needs to invert its risk vs reward dynamics - instead of riskier & more complicated moves being more rewarding, they should be less rewarding. To do this visually, you need to emphasise the more important scoring elements visually while making the less important ones more subtle.
The idea is to give the player a simple, easy to understand game plan upfront and then let them discover additional nuances and complications at their own pace.