MP EV MVP Retro

What I’ve learned from breaking my own rule.

I remade a game again

EVMVMVP is the awkward initialism for my latest project; a multiplayer EV-like. You can download it here, and look at the source code here. I know I said I shouldn’t do it, but I really wanted to learn Godot and having a multiplayer server with multiple levels seemed like a really interesting problem to tackle. Implementing a bunch of stuff I already know pretty well but in a new engine and language has been a fun ride. I’ve jotted some notes down about the project so far. I’ve saved a deep technical discussion of Godot for another post.

Using an engine is well worth it

After flythrough.space I decided that I wasn’t going to build too much from scratch next time, not was I going to fight against tooling. The options I considered for FTS where Panda3D and BabylonJS. However over the course of that project also tried out Unity and Godot for small side demos and ended up liking Godot (and it’s permissive license and thriving community) enough to go all-in. In retrospect this was the absolute right move and I regret nothing.

Working With Onyx’s Sprites

I’ve been staring at these sprites on and off for something like fifteen years, itching to put them into a game. The original models used to render them are long gone, but Onyx (in a tutorial) explained how to separate out the engine glows. Working with these sprites feels like stepping out of time. They’re from the 90s. There are dithering tricks here I wouldn’t even imagine. They use a tiny palette but they’re so damn weighty. Present. Metallic when they need to be, plastic when they want to be. They tend towards the Star Wars / Used Future look. They invite stories to be written about them. And the planets are just absolutely lush too. Something about that limited palette…

Testing with real people is essential

I’ve been nagging more people to play this one. On the one hand, it’s much easier to convince someone to play your game if you can play it with them. On the other hand, you really need to do that testing or you will (like I did) find out that the sync method you’re using is totally busted and you need to replace it with interpolation. Furthermore, you’ve got no idea if your game is fun to anyone but yourself unless you subject it to ruthless testing by people who will give honest feedback. It becomes very easy to get settled in and work around horrible UI issues you’ve coded because “well, I know how it works.”

The ability to test solo is essential

Wrangling people for a test is hard. You need to be able to fire up an instance of your server and client with minimal effort or you’re going to wait too long and leave a bunch of bugs that you’ll be scrambling to fix when you finally get people around to test with. And absolutely nothing is more frustrating than the stars aligning for a test only to have a showstopper rear its head at the worst possible moment.

Procedural Generation

I had a conversation with a coworker about EV games and roguelikes. They made a claim that I’m not sure is true but definitely opened my eyes-they thought that the star maps in EV games probably had been procedurally generated-once.

Actually hand-crafting a full EV map is a bit of a process. One thing I never finished doing in FTS was populating all of the systems.

Meanwhile, Mag Steel glass has developed a really cool universe generator spreadsheet. I figured I could leverage it to make a map. Traditional EV games require you to specify exactly who spawns in what star system and FTS carried on that painful tradition. For this one I had simpler rules; factions had starting locations and then grew “influence” where planets (or stations) would belong to them and their ships would spawn. If two areas of influence overlapped, they’d fight.

Document enough for people to get up and running locally

Port forwarding is a lost art. There was much grumbling when my friends and I went to go start a game of Valheim; luckily I’d learned to properly port forward as part of this project! If you need a server for people to play your game, you can’t rely on your master server; you’ve got to give people an option to join localhost and put it in your manual/readme. Having images in the readme also helps show people what your project is about; this is especially important for games.

You don’t get cross platform for free

Using cross platform technology does not absolve you from cross platform testing. All of your target platforms might have the same features but they may all have different bugs for you to work around. I’m developing on linux out of what is probably just bloody mindedness at this point, so when I did my live demos on windows I often learned exciting new things like ‘your sound exports are busted’ or ‘only one computer can be assigned to a given port by a given router’ or ‘you need to version bump the editor for correct mac exports.’ Guess which one of those lead to an incredibly embarrassing platform bug report!

Video Tutorials are worth it

Part of my commitment to overcoming my stubborn hangups was relenting and watching video tutorials. I still strongly prefer written tutorials as a medium but the content is moving to video. For example:

This (and the followup on extrapolation) gave me enough to totally re implement my networked movement.

Working In Public

I post progress updates on discord channels full of fans from the old days, and regularly demo the game with friends. This has been a source of motivation and helped me stick to a more iteration focused (and less “well, I’ll go off for a month and make all of the models/textures/features”) work habit. I’m not sure I’m going to do this for every project; part of why I did it for this one was I wanted to spur participation and that didn’t really happen. Very few people who I didn’t know personally went to the trouble of getting the game running. Part of that is because it wasn’t immediately approachable (you need to host or find a server then run another client to join it) and wasn’t immediately grippy (ok, I can buy a ship, now what?) Any amount of friction is going to turn away potential players, and if you’re used to playing rough unfinished demos this can become a blind spot. But beyond even that, getting someone to try something new is just plain hard.

People respond to Video

At the end of the day, pictures and text don’t cut it for sharing a game. Games are animation. People are used to watching streamers. You need a video of your thing and it needs to show off what the thing is about, including how you interact with it and why it’s compelling. This was my attempt:

Have features and fixes gone in since that video was made? Sure. But it’s still the best tool to show off what the project is about.

Parting Shots

Has the project met its goals so far?

  • Learn the Godot engine: ✔️

I’m comfortable with the technology now; expect future prototypes to be more rapid.

  • Use Onyx’s Sprites in a game: ✔️

I used most of the ship sprites themselves and lovingly separated out the engine glows for even the ones I didn’t use, in case someone in the future wants to use them again. Cosmic Frontier mod anyone?

  • Build a reusable platform for multiplayer EV-clones: ✔️

Maybe this is a partial check. If someone has an idea and that idea is based on multiplayer EV gameplay, I’m confident that if they’re willing to learn Godot they could at the very least copypasta big chunks of the MPEVMVP codebase and get something running. Or they could fork it, replace everything, add missions, remove multiplayer, and make an EV clone real fast.

  • Discover if EV gameplay works in multiplayer: ✔️

Learned important lessons about how much bullet hell spam a networked game can handle, and as a straight dogfighting sim it let me figure out what is and isn’t fun in PVP EV.

  • Build a game people want/like/play: ❌

As far as I can tell, nobody enjoyed playing this besides on demo night. The service I think people want this game to be is beyond the scope of what it can be with a team this size.

4/5 ain’t bad. Thanks a bunch to everyone who tested it, gave feedback, and flew too far to find the center of the map again.

Flythrough.Space Retrospective

I’m not going to rehash don’t remake an old game, but the principle very much applies. Flythrough.Space is a 3d remake of the EV series, and as such suffers from the “remastering someone else’s game” problem. That said, it was a long and involved software project, so there’s got to be something to take away from it, right?



Content Design

Don’t do content design up front. I had reams and reams of content for this project, a holdover from an even older project, making the oldest data seven years old! And the design of the map dates back to an EV Nova TC I attempted in 2005 or 6. The first plan involved using the graphics from Onyx’s Old Shipyard, back when the whole thing was going to be sprite based. FTS, while using the existing data, was 3d from the beginning so needed its own entire suite of assets.

I ended up using Blender because it’s Free Software and I wanted to keep the toolchain as free as possible. I wanted anyone to be able to pick up the same tools and build more content. It took me a while to become minimally proficient with Blender. One thing that ended up looking really cool was making rough textures with blender’s texture paint then applying a “paint” layer in GIMP with clean lines but dusted up with a rough brush and a makes for a worn look. The ships aren’t exactly up to modern standards, but I’m proud of the results nonetheless.

The planets turned out pretty good. Never did nail clouds though.

inshes.png

Technology

Flythrough.Space started out with very basic requirements. I wanted to do the game in 3d, I wanted menus, I wanted to be able to write object oriented code and I wanted to be able to break code into multiple files. Web was a platform I’d been working on for my whole career up to that point, and I wanted to make the game very easy to pick up and play, so web it was. There was a long period of using babel and Angular to compensate for the shortcomings of JS, but eventually we got all the goodies right in the browser! Once I was working in vanilla ES6 with no precompiler or other baggage, working with the code was pretty much a joy.

ECS

I tried to build an entity/component system. All of the game’s behavior is implemented in “systems” which are functions that get run each frame, requesting entities that have specific components to work on. For example:

export function velocitySystem(entMan){
  // Provides inertia.
  for (let ent of entMan.get_with(['velocity', 'position'])) {
    // Do physics
  }
}

The game state all belongs to entities or the player save object. This turned out to be a fairly awkward way to program a game. It was very nice that, say, velocity or AI could be added to any entity pretty easily. But there just weren’t that many different classes of things in FTS. A class hierarchy wouldn’t have become overly complicated, considering the small number of different entity types in the final game (ship, beam, planet, asteroid, projectile, explosion.) One major drawback was how it changed the code layout of the project. If you want to see what code controls the behavior of a ship, you need to browse several different files and it’s extremely unintuitive. Likewise, if you want to know what parameters a ship can have, there is similarly no single place to check out. There is no FTS bible. One possible evolution would be to put every possible ship param into the implicit base class, but then you’ve got a few parameters where absence matters…

Collisions

I wrote a primitive collision library myself, and my attempt to integrate a proper one later was frustrated by a lack of the primitives I needed, specifically objects with no rotation components.

The crowning achievement of the collision system is the line segment collider. Managing to implement (not invent, mind you, merely implement) allowed continuous collision detection which made combat work much better. It also enabled beams, which are a staple of the genre.

Missions

The missions system leverages ES6’s actually very nice text interpolation to make nice templated text. It uses eval all over the place to do it. It’s a bit fiendish how it works, and it works very nicely for stuff like “deliver cargo” but to add features like “go blow up ship” you’d need to add something in the system code to spawn ships depending on active missions. Probably wouldn’t even be so hard, but it’s low hanging fruit that I haven’t picked. Missions where a very late addition, right before the Alpha release. The json to define a mission looks like this:

  "smuggle": {
    "short_name": "Smuggle ${cgo(this.cargo.type)} to ${this.dest.spob}",
    "desc": "Some shady exporters need ${this.cargo.amount} tons of ${cgo(this.cargo.type)} moved discretely.",
    "offer_if": "true",
    "offer_state": "missions",
    "cargo": {
      "type": "${get_illegal_cargo()}",
      "amount": "randint(5,10)"
    },
    "dest": "get_spob_same_govt()",
    "reward": 14000,
    "accept_modal": {
      "text": "The dockhands quietly load the mislabeled containers of \"${cgo(get_legal_cargo())}\" into your hold."
    }
  },

But that sort of thing only covers random cargo missions. I’d designed the world for FTS without giving too much thought to how it would become involved in any sort of story, besides the idea that the Loyal Suns wanted to take everybody over and the Itaskans wanted you to visit every homeworld. This is a problem because in EV games, you’re supposed to play missions to unlock faction ships, and most of the ships in the game (especially the ones I’m most proud of) where faction ships. My inelegant solution: all ships available all the time. No dynamism. Nobody wins or looses. Some people just fight sometimes and if you hail them they’ll give you the slogans of their government. I think that without plotlines, there isn’t much of a hook. There’s certainly no endgame besides grinding up to an Absolver, Nightshade, Capital, Taoiseach, etc.

Data Driven

Every ship, star system, etc is defined in json files. Many previous game attempts had done a lot of work on making a perfect data loader. In this case, I was able to leverage Javascript’s own builtins to make something I’d consider basically perfect (or at least perfect in that it represents an idea taken to its full extent.) You can declare another object as a given object’s prototype and it just works, inheriting values that aren’t overridden. Objects are created on screen by instancing the data (with some enhancement, of course, but it’s all fairly easy to follow.) There’s no intermediary data is what I’m getting at; the javascript objects in the data directly translate to javascript objects in-game. Upgrades work by acting like a diff. Getting all of this to just work was one of the funnest parts of the project and will be the hardest part to let go of.

Community

The babylonjs community, by way of their forum, was extremely helpful. Having real live people who want your project to succeed and can help you use the tools they’ve built is a joy. I would definitely recommend treating community as a critical feature when picking frameworks and libraries.

That said, you also need to cultivate a community around your own project. That’s one angle where I failed in a big way. It isn’t easy to get other people to demo your game. Starting with something very you can get right into, rather than starting a player out as some sort of lowly level one wimp in an RPG might make a lot of sense. I eventually added cheat codes in the form of querystrings (bonus points if you can find them by looking through the source) so players could start in more powerful ships (and so that I could test them without grinding.) But without a large number of testers, tons of bugs slipped through the cracks. It took me way to long to realize that using Control as the fire button simply wasn’t going to work on mac, for example.

If you want to use FTS

I tried to do with FTS what Ambrosia never did with EV: make a totally open system that anyone can pick up, make a game, and truly own that game with. FTS is licensed under the GPL so if you throw in your own assets and release any code changes, you (yes, you!) can make a commercial EV Clone with it. The game scenario lives in the data and assets folders. If anyone has specific questions about how to implement something, I’d be glad to help. But I realized something: nobody is going to mod your game if they don’t love your game. Making a game easy to mod needs to be a secondary concern to making a great game. If you really truly want to build technology rather than a game, I would recommend partnering with someone with a vision for an excellent game, which brings us to our next point.

Putting the cart before the horse

I thought I knew exactly the game I was trying to build from scratch. However I was building it largely in a vacuum and I certainly wasn’t going back to EV Nova to see how the ship movement and turn speed compared. This was probably a mistake. I was so mired in what I could achieve technically that I wasn’t stopping to ask if the game I was building was actually fun or not.

I should have spent the bulk of the development on the basic combat loop, making sure the ship handling was fun, the AI was correct and tight and fun to play against, and the game balance was decent. Instead I focused on creating tons of content and features because I was trying to work from a feature checklist. Feature parity with a twenty year old game does not a compelling game make, as it turns out. On paper, FTS has what it needs, but I feel that it does not live up to that potential.

It’s a mistake I don’t intend to repeat. Future projects I’d like to get in front of people faster, I’d like to be less stuck on design decisions I didn’t even make, and I’d like to have the emotional fortitude to disengage from a project before it takes up half a decade.

In Appreciation: Termux

I’ve been playing with Termux quite a bit in the last couple of days. Being able to do development-any development-while standing on the train is an awesome feeling. I think the site undersells what it is-it’s your compiler in your pocket. It’s ssh in your pocket. I’m redoing my website on my phone (the mobile experience is crap at the moment) by editing the sources in vim, then hosting it locally with Python.

Beware of side effects on default arguments

Check out this quick interactive session:

Python 2.7.12 (default, Jul  1 2016, 15:12:24) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> def f(y, x={}):
...   print x
...   if y:
...     x['y'] = True
... 
>>> f(False)
{}
>>> f(False, {'x': True})
{'x': True}
>>> f(True, {'x': True})
{'x': True}
>>> f(False)
{}
>>> f(True)
{}
>>> f(False)
{'y': True}
>>> f(True, {'x': True})
{'x': True}
>>> f(False)
{'y': True}

What’s wrong with this picture? Notice that when I call f(True) we assign something to the default argument x. When we call f again, it replaces the default x = {} with the previous value of x!

I suppose the moral of the story here is that in order to avoid side effects, you should not mess with arguments you’re passing in, even if they’re just the default arguments.

Python list comprehensions

If, like me, you learned python and programming at the same time, you may have missed out on advanced features that, while awesome in python, won’t carry over to other languages you’ll “graduate” to using. One of these features is a list comprehension. It lets you in a compact (and readable) way write loops that take a list and return a list, and does not require lambda syntax. They let you write this:

list_squared = []
for x in [1, 2, 3, 4, 5]:
    list_squared.append(x ** 2)

like this:

[x ** 2 for x in [1, 2, 3, 4, 5]]

This really shines when you’re transforming and extracting data – you can also stick an ‘if’ at the end, turning this:

odds_squared = []
for x in [1, 2, 3, 4, 5]:
    if x % 2:
        odds_squared.append(x ** 2)

into this

[x ** 2 for x in [1, 2, 3, 4, 5] if x % 2]

It might not be a huge improvement in amount of code (or LOC, since you’ll want to break complex list comprehensions across multiple lines) but it saves you from potential mistakes, typing ‘append’, and (importantly) having to declare and use another variable name. I consider that a win.

If you found this post exciting, generator expressions will blow your mind.