< Back

Battleship in Zero-K

Real-time Strategy

Finding my way to Spring

As a kid I was exposed to a rather random set of strategy games that I'd learn about from friends. The first ones being the Settlers 4 and Empire Earth 1 & 2, followed by Knights of Honor and Company of Heroes. As soon as I started using the internet to look for more games, I found myself the Ground Control, Command and Conquer and Homeworld series. I'd often invite my friends to play these games, but with real-time strategy games the problem was always the same. There was no good AI to play against and playing against each other would be very one-sided.

Around 2010, I stumbled upon the Spring community. Spring is an RTS engine which replicates most of the functionality known from the classic Total Annihilation game in 3D. Multiple games had been built using that engine and could all be played on a common server. I started out playing the most popular game, Balanced Annihilation. A game modeled directly after the original Total Annihilation. A few years passed where I wasn't very active, until I started playing more regularly in 2014. By that time I had switched over to Zero-K. At first it is rather similar, but as you go deeper there are some major design decisions differentiating it from its ancestor. While always able to beat my friends in local games, I started out well below average in Zero-K's online ratings. Still, the tightly-knit online community kept pulling me back in until I gained a better understanding of the game.

Joining the development effort

I started development on Zero-K in 2014 with my first AI. While the existing AI was already much better than what you'd get from most other games, it had a lot of flaws that I wanted to improve upon. My main topics of interest were better understanding of the terrain, using a custom pathfinder and simple terrain analyzer, and an optimized economic planner that would expand the AI's economy while respecting territorial borders. This sparked a race between three different AI developers: lamer, Anarchid and myself. This went on until 2017, when I stopped working on my AI. Lamer is the one of the three who kept maintaining his code, which is now the new standard AI shipped with the game.

Zero-K had been greenlit on Steam in 2014, but wasn't released on Steam until four years later. The reason being that Evolution RTS, a similar game from the Spring community, had already been released on Steam and failed miserably. Not because it was a bad game, but because what worked for the very niche community of a few dozen hardcore RTS fans doesn't directly work for the general public. In 2017, the development focus of Zero-K finally hinted at a possible release. While the existing devs mostly focused on improving usability, I made my contribution by producing a first teaser trailer and some design improvements of the website. The spark that really got me diving into infrastructure development came from a player called Brackman. He told me about the Whole History Rating system, which I implemented to replace the existing Elo rating. If you want to know all the advantages that a Bayesian rating system brings over classic incremental Elo, feel free to watch my presentation. In November, we released the game on itch.io as a smaller scale test. With some help from other users, especially sprang, I made a longer trailer for the upcoming steam release. On April 27, 2018, we finally released the game on Steam. We didn't get quite as much attention as Evolution RTS, but fortunately the feedback has been very positive.


So what is this game that I've been talking about so much? Zero-K is a real-time strategy game that lies somewhere between Total Annihilation and StarCraft. The player controls an army of self-replicating robots. The focus is on direct engagement between players over economy or army building.

What sets Zero-K apart for me is the approach to UI design. Instead of designing the game with a fixed user interface in mind, the interface is constantly improving. Commands are becoming more powerful and units are becoming smarter. If that makes the game too easy, it is adjusted with additional complexity. This removes mechanical actions from the game and leaves the player to make more meaningful choices instead. So while most real-time strategy games demand high mechanical skills from the player, Zero-K values game knowledge much more than clicking fast. Additionally, these UI improvements have allowed Zero-K to adopt unique features such as fully physically simulated interactions between units and terraforming: modification of the terrain during the game.

Units can surround enemies with a single command
Terraform allows passage