Danmaku production process

[matrigs]

How does the actual process of making a danmaku game look like? Not the actual art and music, but just the technical side, bullet patterns, stage layouts etc. I would mostly be interested in cave-games but i guess that any shmup has very similar stages.

I guess that first thing would be design the bullet pattern and choosing hitbox sizes for the bullets. But how does one actually test it. Does the designer actually fire up the pattern and try to dodge it? Or is there any probability math involved, counting free space between bullets, bullet speed etc? When does one know, without being able to finish it himself, that a pattern is in fact doable or that he has to further enhance its difficulty? Is there always a route through the bullets that designers assume the player will take?

Also, how much playtesting is there at all? The fact that so many blind spots, counterstops and glitches happen makes me wonder if it is common to wait for location tests to somehow "sort it out itself".

[ciox]

I can practically guarantee they just test everything by hand and consider a pattern complete when it's as thick as it gets while still being doable.
Testing includes doing it with speed controls so that you can slow down the entire game to 10% and easily find out if a difficult pattern is solvable at normal speed, Yagawa has such speed controls still available in the debug modes of his later games (starting with Ibara I think) and it must save a ton of time.

The vast majority of development time must be playtesting.. and moving an enemy spawn 5 pixels and testing again and so on.

[matrigs]

Testing includes doing it with speed controls so that you can slow down the entire game to 10%

Wow that's actually a pretty simple yet ingenious trick.

[trap15]

Yagawa has such speed controls still available in the debug modes of his later games (starting with Ibara I think) and it must save a ton of time.

Starting with Batrider. I'm sure such features existed in the testing versions of pretty much all CAVE games though.

[KAI]

If I think "this kind of attack would be cool" or "wouldn't it be pretty if this barrage spreads out like this?" I put it right into the program. In short, I get an idea, program it, and check it on the screen. If it's fun, great. If it's not, I try to think of something I can tweak. If I can't, I scrap it. In general I don't think about the difficulty. If I get an idea, I first materialize it.

[n0rtygames]

How does the actual process of making a danmaku game look like? Not the actual art and music, but just the technical side, bullet patterns, stage layouts etc. I would mostly be interested in cave-games but i guess that any shmup has very similar stages.

You might want to hang around the development forum a bit just to get a feel. We tend to talk about these sorts of things while swearing at each other and trying to find bugs and flaws in each others work (But not Rozyrg)

I guess that first thing would be design the bullet pattern and choosing hitbox sizes for the bullets. But how does one actually test it. Does the designer actually fire up the pattern and try to dodge it?

Yes! Exactly that. Sit on the toilet, think up a pattern.. play another game - try to copy it. Then just try to code it as best as possible, then set about dodging.

Or is there any probability math involved, counting free space between bullets, bullet speed etc? When does one know, without being able to finish it himself, that a pattern is in fact doable or that he has to further enhance its difficulty? Is there always a route through the bullets that designers assume the player will take?

If a pattern can be cleared, then it's considered doable. So long as there is at least one way through a deterministic pattern of bullets (meaning there's no true randomness) -- then it's a success. Fun and difficult are other things entirely of course.

If you look at some of the Japanese players doing rythym games - you'd probably think some of those split second timings and hand/eye movements are just beyond impossible. But they're doing them. Generally, if a pattern can be completed with a slowdown debug feature in place and there's a reasonable number of frames for a player to squeeze through a gap -- then it should be possible for a human being to do it at full speed. Keep in mind that 10 frames in a 60fps game is about 0.16 of a second. So if you're an older developer who no longer has the reflexes of a younger gamer -- you can make a reasonable leap of logic based on your own experiences as a gamer and extrapolate possible situations based on a few variables like that.

That's actually a very long time. You could probably hit that within a few milliseconds on a stopwatch if you were paying attention. All it requires is practise

Also, how much playtesting is there at all? The fact that so many blind spots, counterstops and glitches happen makes me wonder if it is common to wait for location tests to somehow "sort it out itself".

Counterstops are a limitation of how high your hardware can count. Generally speaking, the majority of players won't easily counterstop a game - no matter how much they might go on about them existing. Also, something worth keeping in mind regarding counterstops is that there is always going to be a theoretical maximum that any player can achieve - because a game is not infinite - even if it loops endlessly and doesn't have timers in place to ensure that the game eventually comes to an end. Variables have upper limits, they loop, wrap around and so on.

But with that aside, if you assume that in a simplistic game there are 200 items you can pick up, these are your only source of scoring and each one is worth 1 point - the maximum score you can ever get is 200. When you get that 200 points, you've basically gone all zen and beaten the game.

As for glitches, have you ever heard the phrase "It's not a bug, it's a feature?" -- some glitches are intentionally left in because they end up enhancing gameplay and allow for more opportunities to present themselves. I know I've left a few in, because once they're discovered they actually really begin to add to risk/reward gameplay

For example: In my own game, continuing to hover over an exploding enemy that is still in the dying animation and holding down fire will shed point items as though it was still alive. Left it in, because I thought it was cool as it required you to control an area of space on screen.

Kara Throws in Street Fighter are technically a glitch. As is Bunnyhopping in Quakeworld. Both of these are essential skills for mastering the game/character

[Ghegs]

Moving this to Development.

[Giest118]

It's funny. I've made so many bullet patterns, and yet I can't really define my process. I just throw down bullet spawn commands like James Joyce threw words at Ulysses. After that I playtest them and see how much they wreck my shit. Once I determine that the bullet patterns do a generally desirable amount of kicking my ass, I move on to the next one.

[n0rtygames]

It's funny.

No it isn't

I've made so many bullet patterns

We can tell

and yet I can't really define my process.

I can: Actual Psychosis.

I just throw down bullet spawn commands like James Joyce threw words at Ulysses.

Giest118, no-one else can do the things you do!
Giest118, like a bolt of thunder from the blue
Giest118, always fighting off the evil forces
Bringing peace and bullets to all.

...wrong Ulysses?

[nasty_wolverine]

not just for danmaku, but applicable to any software...

while (software < 99.9% complete) {
step 1: come up with a new idea
step 2: prototype
step 3: playtest the hell out of it
if(step 3 < as much fun you thought it would be) {
choice 1: refactor and go back to step 3
choice 2: scrap it and go back to step 1
} software++
} package and rollout;
return fun+profit;

The while loop is set at 99.9% because to the eyes of developer, no software is ever complete
there is always something more s/he wanted to add
s/he just has to stop working on it or keep adding things till eternity

[n0rtygames]

s/he just has to stop working on it or keep adding things till eternity

and then release patches forever.... and ever... and ever...

[matrigs]

lol, my bad for not realising this forum actually has a DEVELOPMENT SECTION. i have had the hardware section and shmups chat bookmarked for years.

one thing seems still misterious for me. so lets assume that you playtest your pattern and it is doable. not to offend anyone, but i guess that most developers aren't hardcore gamers which leads me to the assumption that when one actually manages to solve a pattern, it's pretty average difficulty wise.

so now - what do you do to make a pattern harder. speed? density? and how, if you aren't anymore able to solve it yourself do you know when to stop?

i guess that this might be hit & miss sometimes, just looking at futari 1.01 for example.

[Giest118]

I do play a lot of shmups, which helps a lot in terms of making patterns that are hard, but fair. If you don't have much experience playing these games and you attempt to make a really hard Hibachi-like pattern, you're probably gonna do it wrong.


Five basic things affect how difficult an attack is to dodge:
* Speed
* Density
* Number of components in the attack
* How different each component is
* How many directions the bullets are coming from, and which directions

What I mean by a "component" is essentially how many clearly different things are in a pattern. If a pattern involves simply a bunch of bullets of the same color and speed radiating out of one place with a pretty uniform density, that attack has one component. If they radiate out of two places, then there are two components. If you then add an aimed five-bullet arc into the mix, that's a third component. The more of these there are, the more interesting the attack tends to be, but it also tends to increase the difficulty of the attack in an exponential way.

In theory you could come up with an algorithm for how hard an attack is based on the five things listed above, but with experience you kind of just get a feel for it.

[matrigs]

In theory you could come up with an algorithm for how hard an attack is based on the five things listed above, but with experience you kind of just get a feel for it.

That's what i was thinking about. I thought it might be possible to at least count the free space at any given time

[nasty_wolverine]

You are gonna need analytical geometry for.that.... Good luck..

[n0rtygames]

That's what i was thinking about. I thought it might be possible to at least count the free space at any given time

Honestly, these things aren't actually that helpful in creating a shooter. Just being good at shooters to a competent level generally allows you to make something fun.

I tried all sorts of tools and calculations to help me try and create the game mathematically. Didn't work out. I even wrote a 'heat map' generator so I could look at bullet density during a boss fight. Didn't help at all.

earlier dev shots:



A computer can only calculate numbers. It has no understanding of intensity, emotion or fun. You've got to play your own game to really get a handle for what's good. You've got to let other people test it in private, you've got to accept feedback - you've got to try suggestions from other people. Even if it upsets you or goes against a gameplay idea you had in your head, you've got to be prepared to change and accept that someone might have a simple idea to make your game 200% more fun. Whether it's variables like ship speed, bullet speed - doesn't matter. Take it all on board, try it out and just see what works.



Basically, I think the best advice we can honestly give you is this:

If you're going to start, you've got to go all the way. You've got to make your game your waifu. Be married to your game.

[Giest118]

Calculating free space is a dumb idea for understanding difficulty. Ketsui's bullet patterns aren't all that dense by bullet hell standards, but it's still one of the hardest in Cave's catalog.

[Formless God]

I kinda envy non-mech designs; you simply start conjuring things from magic circles in midair.

But this won't fly with mechs. When designing a mech boss, do you base the bullet patterns on the boss's structure, or design the boss's looks and placement of components based on its attacks?

[null1024]

I kinda envy non-mech designs; you simply start conjuring things from magic circles in midair.

But this won't fly with mechs.

Sure it will. Just fire some mini-mechs into the spots where you want bullets to appear [you could have a large opening on your main machine to create them], and have them just spawn stuff and fly off. And hell, you could get away with having those mini-mechs once spawned teleporting around for more variety in placement.

It's the same effectively, just with a different aesthetic to go with it. Ideally, you'd also want your designs to lean more futuristic if you want the player to accept the teleporting without question.

Hell, if you're really lazy, you could just warp them [either the bullets or the mechanical emitters] in with a suitably sci-fi warp effect.