AVC at GDC '10, Day Four: Spy party!

AVC at GDC '10, Day Four: Spy party!

John Teti

The GDC organizers made a concerted choice this year to ensure that the keynote speeches would be just that—keynotes, and not glorified press conferences. This reassertion of good sense comes in the wake of Satoru Iwata's keynote last year, in which the Nintendo CEO gloated about the Wii's sales figures, admonished developers to be more like Shigeru Miyamoto, announced a new Zelda game, and then ran off the stage flipping the double-bird and screaming, "Woooo! Nintendo rulez! Sony droolz!" At least that's how I remember it.

This year's big keynote was delivered today by Sid Meier, of Civilization fame, and Meier had the courtesy to talk about game development at a game development conference. His topic was the psychology of "GamePlay" (no idea why he capitalized it like that), and while there was nothing earth-shattering, he did share a bunch of entertaining anecdotes about his creative process.

Meier spent a long stretch talking about how his team struggled with the mathematics of battle in Civilization Revolution during the testing phase. He presented the sample case of a battle in which the player's unit has a strength of 3 and the enemy has strength of 1. In a straight 3-to-1 situation, the player would usually win, but about 25 percent of the time, the computer opponent would beat the odds. Makes sense, right? Yet the testers freaking hated this—although, Meier noted, they did not mind when they won a 1-vs.-3 battle, as they chalked it up to their own "brilliant strategy and tactics."

The development team added more "dice rolls" to each battle and rebalanced the game so that in a 3-to-1 situation, the player was even more likely to regress to the average win. They also found that while a 3-to-1 upset was deemed ludicrous, testers could tolerate the occasional loss when they had a 2-to-1 advantage. "Occasional" being the operative word. If by chance, testers lost two 2-to-1 battles in a row, they were enraged all over again. Once, they could see, but how could it possibly happen a SECOND time after that? So the engineers tweaked the game engine to take earlier battles into account when calculating results. In other words, they baked the gambler's fallacy right into the code.

Then, yet another issue cropped up. While players accepted a loss with a 2-vs.-1 strength advantage, they found a loss with a 20-vs.-10 advantage unconscionable. The game engine considered them the same, though. So again, the mechanics of the game were tweaked to make victory more likely for 20-to-10 than 2-to-1. Through all this, the fundamental balance and difficulty didn't really change. The game was just as challenging as it ever was. The developers were simply futzing with the numbers so that the player accepted the reality being presented on screen—a sort of emotional math. This is what Meier means by the psychology of gameplay. It's not always rational, and developers have to accept that.

Meier also talked about artificial intelligence, and he said that he doesn't try to make his AI act like a person. Even if you make the AI somewhat clever, Meier said, players still perceive it as dumb, because they view themselves as inherently smarter than the computer. And if you really push it—make the AI very smart—players assume that it is "peeking behind the curtain" and cheating. (Puzzle Quest, anyone?) So instead of creating virtual humans, Meier treats the AI as a feedback mechanism against which players can measure their progress. The rival civilizations play with straightforward tactics, presenting a challenge that the player can readily observe and understand. Meier's not trying to outwit players; he's making them refine their strategies and think about the game in new ways.

Along the same lines, Meier said that you have to "protect players from themselves." In testing Civilization Revolution, he saw that players were often saving right before a battle and, if they lost, reloading the save file to try the same battle again until they won. To circumvent this, the developers updated the save system to include the randomly generated battle outcome in the game save—essentially storing the dice rolls. So players who try the save-reload cheat tactic in Civ Rev will find that every time they reload after a battle, they get the same result. This might seem a bit paternalistic at first glance—it's my game, let me play how I want!—but I think it's smart. Meier was removing a source of temptation to preserve the overall fun of his world—"the player can't ruin the game" with this type of approach, he said.

Later, Gus Mastrapa and I joined independent developer Chris Hecker (mentioned in yesterday's writeup) to playtest a two-player game he's working on called SpyParty. Here's the premise: One person plays as a covert spy at a cocktail party with about a dozen other guests—non-player characters that the computer controls. The spy mingles among the NPC guests and completes certain tasks—he (or she) has to plant a bug on the body of an ambassador, swap out a small sculpture with a booby-trapped substitute, etc.

Meanwhile, the second player takes on a "sniper" role. As the sniper, your job is to shoot the spy with the single bullet in your rifle. You watch the party through the windows and try to figure out which of the characters inside is the one being played by a real person—i.e., which one is the spy. As Hecker put it, it's a "reverse Turing test." The person playing the spy tries to move and act as much like the A.I. characters as possible. But the spy still has a set of missions to complete, so he has to risk exposing his identity when, say, his hand swipes the bug onto the pants pocket of the ambassador.

I don't want to beat around the bush here; this game is awesome in too many ways to describe here. The spy player can see the sniper's aiming laser, which ratchets up the tension. When you're playing the spy and you see the red laser drift across your head—meaning that the sniper is keeping a close eye on you at that moment—you have to calm yourself down: Act natural, act natural. Any jerky movement or other human slip-up will expose you. And while the sniper role sounds easy, it's not, because of how many people are milling about the small room. You have to process a ton of visual information, trying to sift through it for any evidence of non-computerized tendencies. (Gus and I both guessed wrong many times in the sniper role.) We would have played all day.

The terrible news is that, according to Hecker, this game is about two years from completion. I don't know if I can wait that long. It's one of those ideas that is so good, yet so basic, that you can't believe nobody ever thought of it before.

Odds and Ends:

- I tried the PlayStation Move. It takes some getting used to. The Wiimote is a pointing device, which means that you can move it a little to make the on-screen cursor move a lot. But the PlayStation Move takes a picture and maps the glowing wand onto the screen image—it's not a pointer. This means that instead of magnifying your motions, the Move actually shrinks them (unless you have a truly enormous TV). The upshot is that if you've ever used a Wii, the Move feels really slow, like you're moving through mud. My guess is that this isn't a fundamental problem—that players will become accustomed to it—but I don't know. My two minutes of play time were mildly amusing but also frustrating.

- Members of the press receive a ton of press releases before and after GDC, and nothing makes my eyes glaze over faster than "New MMO!" announcements. I've tried World Of Warcraft, but I eventually tired of it because it felt like I was doing a lot of watching myself, with occasional keyboard input, and not a lot of playing. I have to admit, though, that I saw an upcoming MMO called Tera today and it unglazed my eyes a bit. The combat system behaves more like an action game than your typical RPG—position matters, and enemies have movement patterns akin to a Zelda boss. You have to read and react to their "tells" in real time. Also, I know this is heresy to some, but you can play the game with a gamepad (the demo I played had an Xbox 360 controller), and I prefer that for a third-person action game. It's being pitched as a "next-generation MMO," which aside from being meaningless is also a bit much, but Tera is different enough that I want to keep an eye on it.

- I forgot to mention a nice line from the IGF Awards last night. The hosts put on 3-D glasses and, after a bit of shtick, observed, "Everything looks the same, but a little bit worse." It's the best summation of 3-D gaming that I've heard, matching my experiences at tech demonstrations on the show floor. There's depth to these gaudy images, but I can't imagine playing an entire game like that.

- Reminder if you are in San Francisco: "Comedy in Games" panel tomorrow, Room 134, North Hall, 10:30 a.m., with Tim Schafer, Rhianna Pratchett, Sean Vanaman, and me. I don't know what the show organizers have planned in terms of video recording, transcript, etc. I will record the audio of the session myself and transcribe it if necessary. Thanks to all of you who have expressed interest.


David Wolinsky

Well, folks, our ride at GDC is nearly at an end. This will be my last filing from San Francisco before getting on a plane early Saturday to head back to Chicago, and somehow it was fitting that this city’s usually accommodating climes did a complete 180 and gave way to unrelenting humidity and rain pissing from the skies non-stop because my entire day at the Moscone Center was noticeably devoid of the word “indie,” something totally unthinkable earlier in the week. I also sampled Tera and attended Meier’s keynote, but my day differed from John’s in that I broke off to hear lectures on two of last year’s biggest titles: Batman: Arkham Asylum and inFamous.

Rocksteady art director David Hego led an understated analysis on how you reboot a troubled superhero game series while also staying true to the source material. Or, at least that’s what the program implied he’d be talking about. Those points were touched on, but instead Hego focused largely on Arkham Asylum’s art direction and lighting, something just as crucial to the moody, dark, and astonishingly great game. (Hey, there’s a reason we voted it the No. 1 game of 2009.) It’s common knowledge that until Arkham Asylum there hasn’t been a single Batman game worthy of the license, and while Hego did acknowledge the Dark Knight’s lengthy history in games, he noticeably didn’t state an opinion one way or another about those previous attempts.

In fact, based on his talk, you got the sense that they and the movies were ignored outright as they were interpretations of the source material: the comics dating back to the late ‘30s. “Great characters come with great expectations,” Hego said, so they did their homework and researched as much as possible on the character and his world to “really understand what Batman is.” He’s a man of the shadows, but also requires light to seize upon his prey, and the “visual narration,” or telling the story and directing the player to follow it subliminally was something they meticulously tinkered with until they got it just right. The challenge, Hego said, came from keeping the game dark, but not monochromatic—to be colorful within the confines of the gothic mood. Similarly, it was important that Arkham Asylum’s 40 rooms and 34 corridors (yup, that’s all there is in the entire game) all didn’t feel the same: “If you go on holiday to a new place, after the first couple of days you stop paying attention to where you are, because you’re just used to it,” so painstaking efforts were taken to assure each of the asylum’s different buildings didn’t feel alike, and evoked a different feeling from the player. For example, the medical wing was built according to Victorian architectural principles and its color scheme intended to instill horror, whereas the maximum security wing was built in an industrial-gothic style meant to make players feel as though they were getting claustrophobic in a bunker. It’s subtle when you play the game, but seemed painfully obvious in Hego’s slides.

The same talk of subliminal ways to craft the player’s experience carried over into Sucker Punch development director Chris Zimmerman’s session on inFamous’ controls. (Another game in our top 10 of last year.) It’s another superhero game, but one based on a character of the company’s own invention. If you haven’t played it, you basically run around and scale buildings as a guy with the newfound ability to shoot lightning from his hands, drift in the sky from surface to surface, and are given the option to zap innocents if it tickles your fancy. There’s a morality system wedged into the game tied largely onto that last part of the game, something Zimmerman joked “wasn’t his fault,” but that had little bearing on the topic at hand. His thesis was this: Super Mario 64 has tricked us all into thinking we can accurately control onscreen characters in three dimensions. inFamous is packed with frenzied firefights, the necessity to float and shoot, and also hopping from surface to surface to survive. Zimmerman tested his fellow staffers when development started by simply asking them to point at onscreen objects with their joystick and then also to hit a button when a ball bounced. He found that, surprisingly, humans are terrible at this: At the very least, the best of us are at least .03 seconds off and have flawed spatial recognition.

That presented a challenge for a game that would require you to vault from the top of a building to the top of a fence and then, say, land on top of a moving train—how can players “seamlessly express their wishes in the game space, be challenged, but also succeed” without hidden aid? The answer took them a year and a half to figure out, and was essentially a series of compromises between assisting the player entirely too much (thereby removing the challenge) and not assisting them at all (thereby removing all chances for success). Basically, gamers on some level are cognizant of the fact that they’re playing a game—but we subconsciously expect a little give in a game’s logic. We want to be treated like rock stars but not be expected to play guitar.

Zimmerman explained how they refined the combat in finding inFamous’ eventual happy medium: “Players are shooting at the last frame, not the current frame,” so you have to allow for some “magic slop,” as he calls it to correct for our natural delayed reactions. If the reticle, which is controlled by the camera, is at all over an enemy when you press the button, then it counts as a hit. But to avoid too much “magic slop,” players expect more leeway on a close-range shot than a long-range shot. If you make a headshot on an enemy 100 miles away, it removes any implication of letting the player feel skilled. But combat was easier to figure out than what most of the one and a half years were “wasted” on: Moving from surface to surface.

Getting back to the initial experiment, Zimmerman played a couple of movies showing how the game infers the player’s movements. A red arrow showed the actual direction being pushed on the joystick, while a white arrow showed the inferred direction reflected in the action onscreen. These movies were literally a couple of seconds, but it was stupefying how much was correction was going on in these clips.

Not even close.

This is probably what’s under the hoods in a lot of games like Uncharted and Assassin’s Creed—which Zimmerman says they tried aping at first, to no avail since inFamous’ landscapes are “denser”—but man, does it say a lot our inability to accurately perceive three dimensions on a TV screen. Presumably it'll only get worse and need more correction as this motion-sensing stuff starts getting more off the ground, but who the hell knows.

Sadly, I'll be missing John's panel on Saturday morning, so again: If you're in San Francisco and can make it, don't miss it. After so much talk about getting inside the psychology of games and indie game-makers rebelling against the wisdom of the previous generation (as they should), it'll be a nice change of pace to have a chunk of time set aside for exploring what games are: fun.