Creating the mid-air battle arena ballet called Rockets Rockets Rockets

Steambirds developer Andy Moore has taken to the sky again, this time soaring even higher and faster, flying past Greenlight's red tape, with Rockets Rockets Rockets. Andy has teamed up with artist Loren Bednar and programmer Kimberly Voll, with Power Up Audio (TowerFall Ascenion, Crypt of the Necrodancer) working on sound for this local- and online-multiplayer dueling ship game hitting Steam Early Access on May 2. Here, the team discusses Rockets' split origin, balance of realism and fantasy, core gameplay, quick solution for online multiplayer, and AI design.

The Rockets Rockets Rockets' core

The one-on-one mode begins with two rockets on their home launch pads. The first one to hit their opponent and score a three-point-lead, and hold it, wins.

While it sounds simple, the chaos that ensues in the above trailer indicates anything but. Not present in the video are the planned hazards, dynamic environments, weapon customizations, power-ups, and other trimmings. Scrapped modes such as capture-the-flag, first-to-three, timed games, and king-of-the-hill are not in the game now, but their code is, meaning the developers could switch them on later.

The gameplay the team is focused on showcasing for now is the human-on-human 1-on-1 couch action, which is what Andy considers the core gameplay. "It'll be present in everything, in some similar form, so it's the easiest thing to show, test, and balance for us. We fully intend to have single play campaigns, practice modes, AI challenges, endless modes, team play, co-op, and online multiplayer.

The versus mechanic even lends itself to the planned Zen-mode. "The rockets perform this dance with each other - we call it 'the dance' when we're playing - and it's hypnotic and suggestive, and some have even said it has romantic tension." Andy said the mode would allow you to paint pictures, change trail colors, take screenshots, and more.

While Zen mode won't have any puzzles or tracing, a racing mode might give rise to such patterns. "Think connect-the-dots," Andy teases.

Online multiplayer? That only took a week

With more and more multiplayer games popping up local-only, this rocket-paced battleground seems like it would follow the same route. However, Andy shares that the online element took only a week to get working.

"Early on we wanted to test out the game's capability of going online before we pursued it, and we ran into what I call the 'TowerFall' problem, which requires instant precision on frame-data because the world can change very quickly.

"We tested out a few assumptions with the netcode and found a clever work-around that allows us to keep the intense, fast-paced and responsive controls, yet still allowing for online play. We're not ready to reveal it yet - we want to get it feeling really really good first, and make a big Early Access update out of it."



That clever workaround required a large rethink on how online games of this type work, and it may not work with the more popular couch games out now. "It changes the fundamental core of how the game plays and feels. You have to build a game from the ground-up with an idea like this... It's not magical code or anything, it's just thinking a bit outside the box on the core mechanics of the game, making sure that inputs work with low latency and don't have adverse effects."

The adverse effect of too many players is something the team is also keen to avoid. They have noticed the best, most exciting matches are 1 on 1 and will probably cap the game at 4 players for balance reasons. "The more players you add the less we can dramatically zoom the camera, so things feel a lot slower and less exciting. That's why we're excited about online multiplayer - multiple screens, multiple cameras to zoom around, and way more clean than doing a splitscreen thing."

Stanley: Kimberly Voll's human-like AI

Kimberly Voll is excited by the challenges Rockets Rockets Rockets' AI presents. She says that AI can be split into two types: "the systems that do things consistently better than humans, and those that do things consistently in human-like ways."

She feels the core of humanity, like the game, rests with the latter. "Our brains are naturally wired to give the benefit of the doubt and see human-like behaviour in things. As AI designers, we need to learn to leverage that and adopt a sort of "less is more" approach, allowing the brain to fill in the blanks for us. Because if we try to do too much, we hit the uncanny valley--that point at which the limits of the portrayed realism and the realism expected differ, or what I like to call when humans call 'BS'." For Rockets, this translates to believable fly patterns, pursuit and evasion behaviours and appropriate failure points.

The AI that Kimberly is working on now is named 'Stanley' as a tribute to the Stanley Parable. "As a part of the program, Stanley, like any AI, has access to far more information than you or I as players of the game, yet his ability to weave meaning into that data is completely at the mercy of the code that I write. He can't use his 'intuitions,' or take in a holistic view of the rocket arena like you or I can with brains dedicated to processing, filtering, and prioritizing the world around us into patterns. Any piece of information that Stanley "considers" must be programmed, as well as the implications that has for his behaviour."

Kimberly wants to create an AI that is challenging on multiple levels. "Stanley v1.0 for the early access release will be a strong flyer, moderately aggressive and hopefully semi-intelligent in the face of many kinds of attacks. As I continue to work on him, you'll start to see his personality emerge. He'll have some favourites, a few signature moves, and a play style ultimately all his own."

The origin of ROCKETS*INFINITY, and the sweet spot for realism

The game whose name was too long for any database and eventually shortened to just three "Rockets" was created by mashing ideas from team members Loren Bednar, Kimberly Voll, and Andy Moore. Andy says he can trace the game directly back to Luftrauser, the original game by Vlambeer. "Back in 2012 I went to TIGJam and so loved Luftrauser, I'm a pilot after all, that I decided to make something similar; I ended up making Project Corona.

"Project Corona was a passion of mine, but I couldn't make it 'click' - there was something missing from it that I couldn't figure out. When Loren came to me with the idea for side-scrolling rocket combat, I figured my old prototype would be a good fit."

Andy starts most of prototypes in the realm of realism. One of his first iterations of Rockets was him trying to model fluid dynamics. "I LOVE realism in games, but it's really hard to make reality FUN." He continually trims away the reality until he arrives at a fun game.

"The reason Project Corona didn't ever take off was because of my too-realistic modeling of the world around it; the more I reduced gravity, the more I reduced the air, the more I reduced the control surface modeling, the more fun it got, the more responsive it felt. Real airplanes and rockets have horrible responsiveness and aren't fun to fly."

Realism isn't all bad, though, as Andy explains, "If you truly make up EVERYTHING in a game, you have to re-train your audience on every single affordance. This is why physics games are easily understandable by even children: they have a ton of practice throwing their food on the floor, over and over again. We know how the system works.

If you present a system that is entirely alien, it's going to come with a LOT of baggage that may be impossible to overcome."

He says that what's interesting for him is taking an existing system that people know how to interact with, and giving them a novel situation to experiment in and discover new solutions. "Rockets was designed by me to just be fun to fly and swoop around, but Kim Voll put a lot of work into making sure there was a large depth of mechanics you can learn and iterate on, improve your abilities, discover new things your rocket can do and map those onto your consciousness

But when you first pick it up: You understand that the arrow-shaped-rocket flies forward when flames shoot out the back, and what goes up must come down. Plus shooting. Easy to understand, hard to master."

There's a bit of realism in the game's mechanics, though. "The rockets themselves behave in a way that 4th graders can, and recently did, understand - basic Newtonian physics, rockets fly around. It even makes some kind of sense that rockets might have weaponry. When you fire a missile, it deploys out and slowly accelerates forward."

Rocketing past Steam Greenlight, pricing Early Access fairly

Along with a simplified title, Andy's team began showing the game in a basic form. "We went to GDC back when the game only had movement, no combat mechanics. We showed it off on Monday night, and it got such an overwhelmingly good response that we decided to spend most of our GDC working on the game!"

One night, they demoed it to Valve, who was immediately excited and wanted it on the platform. "They were even urging us to launch the game in April, but that wasn't possible."

As for tips to handle Greenlight before Valve makes it go away, Andy says, "Make games. Make tons of games. Meet people, go to all the conferences. Show people your best stuff. After 5 years and about 90 games (that's about where I'm at right now), one of those iterations will hit that sweet spot and the doors will all open for you."

The game that hit did the trick, with just six weeks of development and polish, will start out at $5, increasing gradually during major updates until hitting its final price of $15 or $20. "We look at this like a crowdfunding campaign. This is like Kickstarter, except we are guaranteeing you a minimum product."

Rockets Rockets Rockets will be available for Windows, Mac, and Linux via Steam Early Access on May 2. Xbox One and PS4 versions are currently being discussed.