Most games don’t take eight years to make. But Chris Hecker’s SpyParty is unique. With the tiniest of teams, Hecker designed and made his dream game about subtle human behavior, where a sniper tries to identify a spy at a cocktail party and shoot the spy with a single bullet.
The game launched on Thursday on Steam Early Access with all new art, 21 new animated characters, 10 art venues, and eight missions with a lot of variations. I played it with GamesBeat’s Stephanie Chan yesterday, and we took turns playing the spy, who plays a game of deception, and the sniper, who has to tap the power of perception. I think it’s pretty good.
It’s tricky because the human spy can blend in with other artificial intelligence characters at a party. The spy has to accomplish tasks like planting a bug on an ambassador, while the sniper has to identify and shoot the spy. Now Steam players will get a chance to see whether it’s possible to make a fun game based on observing subtle human behavior, rather than just a game full of action.
“I’ve worked on it full-time for eight years. I worked on it part-time before that, and I don’t even want to do that math,” said Hecker, in an interview with GamesBeat.
We geeked out a bit about artificial intelligence at first, and then we got into the details of the game, its design, and the strategies for playing it. Here’s an edited transcript of our interview.
GamesBeat: How are you feeling about this?
Chris Hecker: In some ways, I feel like I’ve spent a lot of time on this, and in the meantime things like artificial intelligence have become so big.
GamesBeat: You would think that you had AI as an idea in this game from way back.
Hecker: Well, I specifically mean the deep learning kind of AI. Computer game AI is pretty easy. The deep learning stuff went from basically not working at all to, “Now I can talk into my phone and a text message comes out.” Something happened in there. I was at Foo Camp a couple years ago, and I was talking to a guy who was—I can’t remember the name of the package, but he was at Cal, working on one of the big open source packages. He started his image processing PhD when it was like, “Now I’ll write the code to find the mustache.” By the end of his PhD, because it takes six years, you just throw all this data at the thing and something pops out.
He saw that transition, and I was talking to him a lot about what happened and what it felt like. A lot of what happened was, now we just have so much data. It’s the same neural net tech from the ‘90s. We just didn’t have a billion images to train it on. That’s why I get a little creeped out by every time there’s one of those Captcha things on a website. It’s just giving Google more data.
It used to be, if you went to a small town, there would be three stores that sold the same kinds of things, and fourth store would struggle, but you could have a top three. There’d be a big one and a couple of runners-up. Now there’s so many positive feedback loops that you get a Facebook and you get a Google.
GamesBeat: Nvidia shows that video about flower recognition. They show a CPU recognizing flowers, which is pretty slow, and then they turn on the GPU. Suddenly, it’s a screen full of dots.
Hecker: Then there’s these [custom chips] that are custom coded for it. All of Google’s stuff apparently runs on that. It’s not even GPUs anymore. You saw that with bitcoin. The GPU prices went through the roof, but nobody does real mining on graphics processing units (GPUs). It’s all completely custom hardware. We’ll have custom AI hardware.
It’s weird. I feel a bit left behind. I try to stay on top of things. I’m not a fashion-oriented programmer. [laughs] But this feels big enough that I should learn this stuff. SpyParty is a perfect platform to do this kind of thing.
GamesBeat: I don’t think anybody’s gotten to what this means for games yet.
Hecker: To some extent, that’s not even the most interesting question to me. It has such wide-ranging privacy concerns. Stuff way beyond games. Games are entertainment and art. I can think of 20 different places in SpyParty where you could do a deep learning thing, from a testing and debugging standpoint to playability and stuff. But what it means for society as a whole is really interesting. We don’t know that either.
GamesBeat: But SpyParty’s here, so….
Hecker: I have a Steam early access launch date. That’s terrifying. The game has been in development full time for eight years. I got laid off from Maxis at the end of 2009. We’re now at the beginning of 2018. The game was actually started before that, at a game jam. I don’t even want to do that math. But full time it’s been eight years. It’s been for sale on SpyParty.com for about five or six. It’s sold about 24,000 copies, which is great for an indie game, knock on wood.
Hopefully that’s not the sum total of the people who want to play the game. Hopefully there’s more than that once it goes on Steam. Who knows? The indie game business is pretty scary. The fact that I’ve sold 24,000 copies off of Steam, just a Paypal link on my website, is a very good number for an indie game. But does that indicate that Steam will sell well? I have no idea. It’s this weird thing, to be 47 and have a kid and a mortgage in the bay area and not have any idea what my life looks like in 30 days. Do I have to get a job? I have no visibility. That’s very strange. It feels like I’m 18 and going away to college.
I can program a computer and I live in the bay area. I’m not gonna starve. The floor is not that low. I don’t think I’ll be homeless or anything, thankfully. But I’d like to keep working on this game, because it’s coming out really well.
GamesBeat: Tell us what’s new.
Hecker: I have a giant content drop of all the remaining old art converted to new art. You remember what the game used to look like? Here’s Balcony, and this is what it looks like now. Everything is going to die when the Steam version launched, so I need to get all this content debugged in the 10 days between the announcement and April 12.
Everything has been hit. There are four new art levels and six old art levels left in the game right now. The remaining six have been converted. We’re more than doubling the amount of new art in the game. The whole game looks like this now. You can find the old art in there, because the old-timers would kill me if I took it out, but you have to go looking for it, rather than it being the first thing you see. The UI is all fancy and new.
Any time you go more accessible, each step—now there’s a tutorial, for example. Before you had to read a four-page manual and that’s not going to fly on Steam. We’re doing matchmaking instead of a raw lobby. Right now, if you went in there, you’d have to invite someone to play. People don’t like to talk to other people on the internet these days. It’s great, because my community is awesome, but it doesn’t scale to steam. I need to have matchmaking where you can just dip in.
The third thing is I’m raising the price from $15 to $24.99. I wanted to give people a heads-up. A lot of people have been following the game and waiting for it to be on Steam for a long time, so I wanted to say, “Look, I’m gonna raise the price.” The players told me to do it. “Your game’s underpriced at $15, especially now that it looks like this.” From the 2nd to the 12th it’ll still be $15 on SpyParty.com, and then it’ll go to $25 on both.
There’s so much stuff in the game now. We have this insane triple-A level replay system. It’s way over-engineered. We can go through some of those things. This is Balcony. You can see, the layout is exactly the same. It’s just new art. When I originally did this map, I said, “Okay, the game is really tense, so I’ll do a simple map, a small one.” The sniper doesn’t even have to move to see everything. That’s arcade mode SpyParty. But it turns out that doesn’t work. This is one of the most intense competitive maps, because it’s completely behavioral. You have to be incredibly subtle as the spy, because the sniper can see the whole thing all the time.
GamesBeat: What is the spy trying to do, again?
Hecker: The high-level game — there’s a cocktail party, a fancy cocktail party. The spy is trying to do missions, tropes from spy and mystery fiction. Bug the ambassador. Steal the statue. Contact the double agent. Things like that. Meanwhile, looking in from outside is the sniper. It’s an esport-ey competitive multiplayer game, but about subtle human behavior.
I didn’t want to make it about space marines or killing orcs or blowing stuff up. I wanted to see what a competitive game that’s about behavior looks like. Knock on wood, it’s worked out really well so far, because my top players have 20,000 games and more than 1,000 hours. There are tournaments and ladders and all that stuff. I’ve been super lucky. It’s worked out better than I thought it would.
When I first told Will Wright the idea, back when I was on Spore, he said, “That’s not going to work. It’s too easy to see who the AI is and who the human is.” The first thing I prototyped was a thing that had me just walking around a small room with a bunch of other AI, and you couldn’t tell who I was. From there I knew I was right, he was wrong, and this was going to work. The alternative is easy. If the spy is too hard to find, you can make it easier for the spy to be found. But if the spy is too easy to find off the bat, you’re doomed. I’m glad it went the other way.
Each of these maps, aesthetically they’re very different. You can see the different characters. Equal genders. Here’s a character in a sari, a Sikh, a guy in a wheelchair, a dwarf, different races and ages. There are old people. You never see old people in video games. We have multiple gray-haired people. A guy with a cane. We tried to be as diverse as possible with the characters, and also architecturally.
What’s not immediately clear until you play is that the game design is diverse. Each map is exploring a different part of the game design space. For example, the Balcony map I was showing you before, the sniper doesn’t have to move to see the entire thing. This map, the Courtyard, which looks like this if you download the game right now, but now looks like this—that’s Alan Turing in the middle, by the way, because the whole game is a Turing test. A little homage. There are little easter eggs about him all over this map. The reliefs around here, some of them have to do with his computational biology work, and then that’s him in the Enigma machine at the top.
This map has a big statue in the middle, so the sniper is on the outside and rotating around. Now, the spy can see the laser sight that is the center of the sniper’s view. Where they’re aiming, basically. But the screen looks like this, so you could be looking from here and aiming the laser over here. It gives you the general cone of where the sniper’s looking, but you don’t know exactly what they’re looking at. Here, because it’s in the round, the spy can see where the sniper is and can know generally what they can see. This adds a big static occluder in the middle.
GamesBeat: Is that a real place?
Hecker: No, this is a fictional place. The statue is based on a statue that I think is at Bletchley Park or one of the Turing museums. He was such a cool guy, and he got so screwed by the British government, I figured he deserved the recognition.
One of the things I learned from Will, when you’re talking about your game, you should read the room. Describe your game differently depending on who you’re talking to. If I’m talking to a bunch of people who are not in games at all, just normal people at a party or whatever, I might lead with, “Oh, it’s a game about a cocktail party.” Which makes a normal person say, “Wait, a game about a cocktail party? What’s that mean?” Whereas if I’m talking to a programmer, I’ll say, “It’s a reverse Turing test,” and they’ll say, “What’s that mean?” How you get someone interested in your ideas is totally dependent on who that person is and the context they bring to it. That was an important learning from Will during Spore.
Sometimes you make a mistake on that. I just started right in, forgetting how long it’s been since the last time we talked. Wait, rewind, refresh me on the actual game. You can get in too deep. You have to be careful with it. Here’s Gallery, a SoHo art gallery. This map is all about dynamic occlusion. You can see the entire map, but all the people, the guests are in the way all the time. That’s different from the Courtyard map, where it’s static occlusion. Here, the spy needs to wait opportunistically for people to block things. All the missions have subtle tells. You can see the statue swap, or sometimes there’s a deductive audio tell. This map really shows off some of our timeless art style. There’s a Vespa and an old-style cab.