AI War Interview


posted 12/10/2009 by Charles Husemann
other articles by Charles Husemann
One Page Platforms: PC
We’ve seen it in action, but how about telling us what AI War is all about from your point of view. How does the final game differ from your original idea?
I’ve mentioned already that I’m a believer in iterative development, and that really shows with AI War – the original concept is almost nothing like the actual finished game. AI War started out as a turn-based affair, emphasizing line of sight and positioning, with symmetrical AI. However, through extensive, extensive playtesting and brainstorming with my alpha testers, the design shifted over a period of months to what you see now – realtime, asymmetrical, no LoS concerns. We’d play the alpha version twice each week, and then talk about what was going right and what was not. Then I’d work on a new build, scrapping the un-fun parts and adding new ideas for the next week’s session. The result was something that I never could have imagined in advance, and that’s the key benefit of iterative design versus more traditional methods.

However, all that said, there are certain facets of the game – what I like to call my “immutable design goals” – that never changed despite how much else was changing during our iterative process. I knew I wanted this to involve a lot of difficult strategic decisions, that it should play out over 7-12 hours or so per campaign, that co-op was going to be a central feature, that cunning AI was to be another central feature, that it was to have as large a unit cap as possible, and that it had to have no “easy shortcuts” like rushing or turtling in other realtime games. When looking for what AI War is all about, I think that list of goals actually sums it up pretty nicely!

What was the hardest part in developing the game? What did you think was going to be hard but turned out to be simple?
The basics of the engine of AI War were something I had already created for another Arcen title (Alden Ridge, which is slated for 2011), so a lot of the rendering and architecture and general engine concerns were already well-established. What was completely new to me with AI War was the multiplayer networking and AI – and both of those were quite challenging, to be sure. They worked out quite well in the end, but those were the two elements that gave me the most heartburn during development, so to speak. Multiplayer desyncs were the bane of my existence during alpha, as is typical for a lot of RTS developers from what I’ve heard.

Something that I thought would be really hard, but which turned out to be (relatively) simple, was the high unit cap of the game. I have always been a fan of large RTS games, and having a cap of 500 units per player as is common has always felt limiting to me. In AI War, I had fantasies of maybe having a global population cap of 15,000 units for 4-6 players, but I figured I’d have to do a lot of tricky things such as “draw three units but treat it as one,” or something like that – maybe “stacking” of units as in Civ games, or something like that. In the end, I was able to achieve 40,000 to 60,000 units simulated in realtime without lag on upper-middle-end hardware, and that was a huge surprise to me.

I’ve actually recently had a player who is 55 hours into a campaign, with 201,000 units in his map – I was blown away, having never seen more than 90,000 at once. That 200k campaign doesn’t run well on even my quad core workstation, but the player in question has his quad-core overclocked to 3.5 Ghz and says it runs without lag except for a tiny bit when he issues commands to 4,000+ guys at once. I had just assumed that these sort of numbers were absolutely impossible, but for a space game with the right sort of engine and 2D graphics instead of 3D, it’s demonstrably possible – and that’s in the C# programming language, not even C or C++.

Making a multiplayer game with no player vs. player conflict is an interesting choice. How does this team-only approach fit into the AI War universe?
The decision to not include pvp was one that I made for a few reasons I’ve hinted at above. First off, I don’t play pvp RTS games myself, so I’m not an effective SME for them. In order to really make a great pvp experience I’d need someone with lots of experience in that area, but that person isn’t me; I knew just how to set up and design a co-op experience, though, so I put all my energies into that and the single player modes rather than spreading my attention around too thinly.

Secondly, the RTS market is just glutted with competitive strategy titles. If you want to play competitively, the sky is the limit – you can pick most any RTS game that strikes your fancy in a given year, and it will probably do a serviceable job for you at least. To me, not only is that a market in which I don’t have any expertise, it’s also a very crowded market. I’m a huge proponent of co-op in general, and believe that every game should ideally be playable with friends or family (who wants to sit alone every time they want to game – or, worse, who wants to obliterate their lesser-skilled family members every time they try to play them in pvp?). For me, the team-based aspects were my entire purpose in creating the game.
Page 2 of 4