Introversion Explain Subversion's Fate

Earlier this week I had a chance to talk to Introversion lead programmer, Chris Delay. He explained a bit about what was going on with the British indie, talking about how the suspended Subversion project had changed their outlook, and how they're returning to their original approach of being bedroom programmers.

RPS: Ok, let's stop rambling about our various secret projects and get to some material that I can actually print on RPS.

Delay: Ok, let's do that.

RPS: What happened between the Subversion demo that we saw at World Of Love and the announcement of Prison Architect, just a few weeks ago?

Delay: You were there for that?

RPS: Yes.

Delay: Right, well what happened actually started before World Of Love. The demo there was ... well, we set ourselves internal objectives and with Subversion we were aiming for a vertical slice, so a fully playable level of the game. And we hadn't had that, and it had been worrying us. The World Of Love demo was that: a fully playable level, top to bottom. But even as we were doing it we kind of knew that it didn't measure up. It looked good and it sounded good, and stylistically it was nice. But... it didn't play. I would never have let anyone play it, because they would just have said "this sucks." In the case of that level there was only way to beat the bank. If you drew that bank vault it would be "inner door", "outer door", "bank vault". A linear line of doors. There was only one way to complete it! It was quite nice in the sense that it had a bit of time pressure on you, but it had taken us months and months to get to that level. I started to realise that every level in Subversion was going to boil down to this, to breaking a series of locks and stealing the loot.

It's difficult because game design isn't really reductionist. Well, most games do break down to something simple like that, but it wasn't just that which was at fault, there wasn't enough gameplay possibility from the amount of work that we'd done. We had very little in the way of gameplay, but all this visual style and all these tech demos, and I couldn't see how the gameplay was going to come into that, either. That was happening at the same time as the World Of Love demo.

We spent some time to solve this problem, to do an alpha build and put it out, whatever. And eventually we hit on the idea of another game. Prison Architect has a lot in common with Subversion, but I could see the route to the game immediately.

RPS: So... you could see, from the work you'd done on Subversion, a different, but incompatible game dynamic? One that was better?

Delay: Yes. There was a level in Subversion where you were supposed to bust your buddy out of a jail, and we spent months on that level, because prisons are complex places with day-night regimes and so on. We'd invested loads of time in simulating a prison with a few gaps - the places where the cunning player could sneak in or whatever. That level was the gestation of Prison Architect. If you take that stuff and flip it on its head then you have a lot of the systems you need in place for the player to build their prison. That lead to the game we ended up doing. There's a lot in Subversion that is also in Prison Architect, it's just that your interaction with the world is the other way up.

RPS: So you don't feel like you wasted an enormous amount of time and resources on Subversion only to end up doing something else?

Delay: Yeah. Yeah, I do feel that way a little bit. It is quite reckless. It isn't the first instance of this happening at Introversion, it has to be said. We have binned work in the past. The fault here is that Subversion was experimental, right from the very start. Its focus was on procedural generation, and we spent a lot of time developing that. It was always going on the background while we were doing other things. We didn't mind that there wasn't a game on the surface, it must be in there somewhere, right? We've got all this great technology! But it wasn't in there. Subversion is now on hold indefinitely, and that is very wasteful. But it's also fertile ground for other games.

The core idea: hi-tech heist, wasn't working. Take the prison level: blow a hole through the wall! That's it. If you can't make a game out of it, then that's it. Over.

RPS: Hmm, but lots of games are a very simple act repeated over and over. Or they're a toolbox with weak parameters for jeopardy and challenge. People are happy enough with numbers going up! So what was it that meant Subversion had to be canned? Why wasn't the feature you had made enough?

Delay: Sure. You could say that. You could look at Uplink and say "you run a bypass on all the security systems until you've cut them all off", but I think in any game that's any good, the core activity is really satisfying. World Of Warcraft is just pushing the first six number keys over and over, but because of the progression and the visuals and the knowledge of where you are, all the feedback, that's really fun without anything else. The rest of it is theme. In Darwinia you are playing with Cannon Fodder-style controls. You are left and right clicking to control your squaddies, shooting monsters. That would have worked even if it was set in the real world, the core would have been repeatable. Come to Subversion and the core activities weren't any fun. The systems were overly complex. Even the inventory! Rather than have an inventory bag we had this full head to toe system where you could load up backpacks and armour and stuff. You could slot equipment into every part of your body, put bags inside bags, and so on.

And because we'd set it in the real world, during the day, you didn't feel like you were doing anything that exciting. There was no feeling of danger. I can see how you could improve all of those problems and fix a lot of those issues. But the amount of effort we were having to invest was astronomical. And we looked at other games with a simple core mechanic and they were fun right away. Monaco, for example, does a lot of the same stuff that Subversion does, but it's more fun. I looked at that and thought: "he's nailed it, right there." It was just fun to do, over and over, in precisely the way Subversion wasn't. And that happened a number of times with a number of different games, where we saw something that we regarded as core to Subversion, done much better.

Of course if we hadn't had a really strong idea we'd still be battling with all that now. We'd be doing an interview about the difficulties of making Subversion. But other ideas come forward sometimes, and they're much better, and you have to pursue them.

RPS: Understood. And to an extent it feels like the difference between developers who are aware of games other than those they directly inspired by, and those who aren't... I sometimes envy the blinkeredness of some people, because they just get on with what they were doing, no interference, for better or worse...

Delay: Well, that wasn't the primary reason. It was just another factor. It's more the sense that the competition highlighted to me that we were lacking in a core game. We used to be really be blinkered actually. Particularly with Darwinia. We just went ahead and made it. And that worked. But then we suffered from it with Multiwinia, because we made a game that was more like a traditional RTS, but we hadn't accepted many of the standard methods of control, and that gave us quite a lot of trouble.

RPS: Multiwinia was such an ambitious game, because multiplayer RTS is so hard to compete in as a genre... there are only a few contenders.

Delay: We know that now!

RPS: Yes.

Delay: Yes. But you know the history of how we got there. It was all our own decisions that led there, but the decisions were compromised by the fact that we were trying to do a console version of Darwinia for Microsoft. It required multiplayer... But it's in the past! Introversion, for me, was supposed to be a company that made new games. We are supposed to be fearless, making one game after another. The issue was that we made a bit too much money from Defcon, and the temptation to grow into a microstudio with an office and staff took over. We've gone back to the embryonic stage where it's just a bunch of us at home. What we have to worry about now is just our own salaries and working on the next game. Which is always where I wanted to be, I think.

RPS: That's really interesting. What's also interesting is the approach you took with Subversion - procedurally generating a city - and I wonder what you have learned from that?

Delay: I have learned a few things about it! I love procedural generation, and I think I would point at Minecraft as a good example of that. It generates these worlds that are interesting to explore. And that's useful. You see procedural approaches work well for landscapes and cities, at least from that big, helicopter-view angle. It works wonderfully for say, Darwinia, because all the levels there were procedurally generated. Same for Minecraft, the scale is just right. But it's really difficult to apply to complex man-made stuff, because they're designed by someone. Procedural generation can do this stuff, but it's an order of magnitude harder to make it believable.

RPS: Yes. It's like... different sets of expectations and affordances. People are designed for landscapes and react to them in a certain way, rooms are designed for people, and so we approach them, read them, in an entirely different way. Hmm, what am I trying to say... basically you don't walk out into a natural landscape and expect a handrail, but you do at the edge of a building. Something like that. Generating one comes with different interpretations from reading another.

Delay: It's also something to do with the amount of information in a setting, I think. In a heavily designed environment, like a building interior, it's tough to design it reliably, procedurally. You end up with bathrooms in staircases and stuff like that. Human beings look at it and can say "that's totally wrong", but procedural generation can't guarantee that stuff won't happen. Despite all the work we did with Subversion's city generation you'd still end up with a skyscraper in the middle of a roundabout. We struggle to rule it.

RPS: Yes! And that's the thing with doing anything algorithmically. People are happy to assume that the complexity somehow equals intelligence, but it doesn't. It's totally blind. It's still just a machine.

Delay: A machine with a percentage success rate. And in a Subversion city that would be 95% correct, or plausible. And there will be a error here and there. The same is true of Minecraft, but no one cares. The landscapes are error tolerant, because the floating islands look brilliant, and who cares if there's a random magma pit bubbling in the middle of nowhere?

RPS: Haha, well one of the biggest battles we've had internally in my landscape generation project is over the floating islands. One side says that they show the errors, show the process breaking down, show the workings, while the other side says "it's Syd Mead!"

Delay: It's virtually impossible to eliminate. It's bordering on an unsolvable computer science problem, I'm sure it is.

RPS: Yes, I am sure it is. But I think that shows where procedural generation will go. It either ends up being error-tolerant, as in Minecraft, or it becomes a first pass content tool. CCP created the Eve universe by generating it once and then going into prune out the madness and etch in the detail. But you keep the beautiful, natural randomness.

Delay: We did consider that approach for Subversion. We had considered generating loads of banks, throwing away the ones that were broken, and then including the working ones in the shipped game. We could have hundreds in there, and it would be procedural, but human directed.

RPS: Yes, that's really telling. I think that's an insight into game design there. It's almost never down to what is possible, it's down to what is more work and where do you limit the more work.

Delay: Subversion was far too ambitious. Far too ambitious. It was too ambitious to try and generate a full city. The generation task becomes a nightmare. Generating sensible buildings is so hard to do. You end up fiddling with a slider between realistic and tedious. You want game-level layouts, which are virtually impossible to achieve. I think we thought if we could do that stuff, we'd have a great game. But I am no longer sure it would be that interesting. Even if you could create that city, what do you do with it after the first few hours? You need Minecraft's capacity to construct and edit.

RPS: Right, the only way to keep it interesting, probably, is to curate it. And the only way you can get that manpower is probably to crowd-source it? Let the players decide?

Delay: Yes, we had a lot of plans in that area. But then you are into user-generated content and that's also ambitious beyond belief. Even the best companies haven't managed to take full advantage of that. And it's hard work. The game world has to be rich enough for the players to want to spend all this time making stuff, and then you also have to put up with all the giant penises and law-breaking copyright stuff they are going to make.

But yes, all this stuff is indicative of how beyond our reach Subversion was. Now, almost a year after it went on hold, I can see the big picture. And I think there was a really good game in there, in that technology. It's just not like the one we were making.

RPS: So you see a future where you come back to all this?

Delay: I do see a future, but I am not going to promise anything, because I just don't know. But I'd like to come back. I'd like to come back and do stuff with some of the technology that we developed along the way.

I still think a hi-tech heist hasn't been done well yet. But the theme isn't enough. It needs a core game mechanic, and that's not in the theme. And we'd need to find that.

RPS: Thanks for your time.