A quite good day today; three of the five sessions that I attended were extremely interesting. My notes:

9:00am: GDC Microtalks. I was wondering if I should go to it—my experiences yesterday didn’t leave me entirely sold on the value of panels—but I figured that, if I got one interesting thing out of it, I wouldn’t regret going, and probably one of the nine speakers would say something that resonated with me.

In fact, it turned out to be awesome! It was in a pecha kucha format: this meant that the presenters had do do a lot of hard thinking about what little they were going to say and about how to use the slides to best effect; the result was that I got something interesting out of every single one of the nine.


  • John Sharp: I loved the slide showing a Chinese general playing go while simultaneously undergoing bone surgery.
  • Tracy Fullerton: Masterful play can’t be a solitary activity, it needs others both to participate and to appreciate. So you need to develop a culture of mastery.
  • N’Gai Croal: Some interesting examples of player-controlled difficulty. In The World Ends with You, your character levels up, but you can choose to play at a lower level. (Decreasing both difficulty and item drops.) In Gears of War 2 co-op mode, the two players can choose difficulty independently. In God Hand, the game gets more difficult as your combo meter increases.
  • Robin Hunicke: Players want creativity, collection, complexity, community. (I think; my notes aren’t entirely legible.) Players are trying to do all four in Home, but the environment doesn’t support their efforts enough; she had some rather focused suggestions on how to provide more vehicles for such desires.
  • Eric Zimmerman: Spontaneous play; he got the whole audience to play a cooperative game.
  • Clint Hocking: He doesn’t like grade inflation, and doesn’t like 100 point scales where the difference between 89 and 90 is much larger than the difference between 90 and 91. He drew an interesting systems diagram linking these two: the 100 point scale makes it easier to have tiny amounts of inflation each year, which adds up over a decade. So he prefers a five-star scale. Which was interesting (and his slides were great), though I’m not really convinced: for one thing, grade inflation doesn’t bother me; for another reason, the term “grade inflation” comes from a five-star scale, namely A/B/C/D/F.
  • Jenova Chen: He had a pie chart dividing experiences into intellectual, emotional, and social, with some interesting examples of art forms whose sweet spot is at various places within one of those or on the border. Games are still pretty weak on the social side.
  • Frank Lantz: Started of with saying that “games aren’t a form of media”. And then he proceeded to justify it by starting with statements that made me roll my eyes but somehow transitioned to statements that made a lot more sense; I wish I’d caught the transition from the one to the other.
  • Jane McGonigal. CZADOF: Confucius, the Zombie Apocalypse, Dance Off, and Fractions. All things we should strive for in games; the Confucius bit was related to a notion of “jen”, which means something like kindness.

10:30am: Kojima’s keynote. Mostly banal: you can climb over an impossible game design wall through a combination of improved hardware, improved software, and design. Slightly less banal: you can also overcome it by walking around the wall, climbing a different (and lower) wall. Subtext that’s particularly interesting in today’s landscape: Mario can jump higher than any of us.

1:30pm: Creating Replayable Cooperative Experiences, basically a lot of discussion of Left 4 Dead design decisions; I now have an overwhelming desire to go out and play that game some more. Some notes:

  • Successful coop games have legs; with L4D, Valve was trying to design a game where playing cooperatively is the only successful strategy. So, from a design point of view, they treat the entire team as a unit, and make sure that, if one player becomes isolated, that player will die. But they want to do this without being too heavy-handed.
  • Zombie movies provide an familiar framework whose conventions reinforce the importance of cooperation. (E.g. jerks go off alone and die a horrible death.)
  • The special infected all exist to solve specific problems, as well as adding variety. The hunter will kill off stragglers/lone wolf players. The smoker will also do that, but will also add chaos to tightly coordinated teams that are at risk of being so good that the gameplay gets dull, because all of a sudden one team member gets dragged far away. The boomer teaches you to be careful where you shoot, and sets up some dramatic anticipation (about which more later).
  • Special infected’s incapacitating attacks, besides creating a fear of separation, also give characters an opportunity to be the hero; also, helping is fun! Ditto for other situations where people are helpless/incapacitated (e.g. hanging from a ledge, or health dropped to zero).
  • And then there are the boss zombies: they give the strongest “oh, shit!” moments in the game, lead to more dramatic anticipation, and make the team talk and re-evaluate their tactics. The tank halts your forward momentum, forces you into a defensive instead of an attacking mode, and its throw makes you re-evaluate your environment; the witch really teaches you to be careful where you shoot (and point your flashlight!), and adds a stealth mode to the gameplay.
  • Characters’ automatic vocalizations in the games have three roles: they provide situational awareness (“The Hunter’s got Zoey!”), they tell you about short-term goals (“The subway is just up the street”), and provide a baseline level of cameraderie (“Thanks for that” after healing a teammate).
  • Limited critical resources might be expected to discourage sharing; in fact, they end up encouraging sharing, because you really need to keep everybody alive. Also, healing other people helps break the ice.
  • They spend a lot of time thinking about dramatic anticipation: key events are signaled in advance, so you have a few seconds when you know that something bad is going to happen and are (pleasantly) panicked. All the special infected have their own noise to signal their presence; in particular, the witch’s cry and the tank’s stomp let you know that bad things are going to happen soon. Or when you get vomited on by a boomer, you know that a horde is going to descend upon you in a few seconds. And then there’s the end of the game, when you’ve made contact with the rescue vehicle: you know that hordes (and tanks) are going to be attacking, and you’re first counting off the seconds until that happens and then the minutes until the vehicle actually arrives.
  • They like structural unpredictability: very dramatic events with low predictability leads to an experience that is always good and that, at its best, can be quite memorable. There are many possible dramatic scenarios (basically all procedurally generated rather than scripted), but only a few will happen in any given run.
  • They strive for alternating peaks and valleys in your emotional experience; their Counter-Strike experience shows that this works well. They start with a baseline scenario that’s randomly generated but that fits this model. And then they add in a model for players’ emotional intensity (updated in response to what actually happens), and use this as a second level of tweaking. This latter leads to a state machine where they’re building up intensity, then sustaining peak intensity for 3-5 seconds, then playing out the current scenario as the peak fades, then relaxing for 30-45 seconds. (Or until you’ve travelled far enough since the last peak.)
  • The bosses aren’t affected by the emotional intensity model, though: they have a different random model for those two zombie types, which they stick to no matter what excitement has happened recently.
  • If you don’t have four players, then you can fill the extra spots by bots. This is useful for three reasons: for one thing, the game designers can assume that there are always four players present, making their task easier. For another thing, the scenarios are long enough that it’s useful to let players drop in and drop out. And the third is an area dear to my heart: they use it for automated testing, by letting the game run with bots overnight to uncover bugs!

3:00pm: Helping Your Players Feel Smart: Puzzles as User Interface, by Randy Smith. Puzzles are great: they let players feel smart, they let players feel actively involved in their progress through the game, and they provide a change in pace in gameplay. Except that they often lead to players feeling dumb: if you don’t help players enough, they can’t solve the puzzle, but if you help players too much, they feel like they must be idiots to be treated that way.

He then presented some notions from user-centered design. Specifically:

  • Visibility: Players should know that key parts to the puzzle exist! This applies to steps in solving a puzzle, not just the objects. Of course, sometimes you’ll want to obscure some of that—these are puzzles, after all.
  • Affordances: Players have notions of what they can and can’t do with an object; if they have the wrong ideas, puzzles will be mystifying, while deducing less-visible affordances can be a key step in solving a puzzle. (Oh, that platform can move!)
  • Visual Language: It helps to have consistent cues as to what objects are interactive and what aren’t, and what their affordances are.
  • Feedback: Communicate what has changed after an action. Or what hasn’t changed: explicit negative feedback is very useful, too.
  • Mapping: Which objects map to what? My notes here aren’t very clear, alas…
  • Conceptual Models: An important part of puzzle solving is coming up with a model of how a system works under the hood.

He then talked about puzzle structures: the steps players go through to solve the puzzle. (Important: a list of steps, not a list of objects.) Combine this with playtest data to improve the quality of your puzzles, in two ways: if people are wandering around aimlessly off of the path given by the puzzle structure, then apply the above notions to make the right path clearer. Whereas if people are actively on a path that is a plausible puzzle structure but not one that will actually solve your puzzle, then make the negative feedback more explicit so they’ll realize that they’re going down an ineffective route. E.g. if there’s a switch at the floor and people are shooting at it instead of putting something on it, then you can increase your negative feedback by having your shots bounce off the switch in a particularly obvious way, and you can also raise the switch to make it more obviously a pressure switch, increasing its depressability affordance.

If you do this well, you get guidance on demand. Some players players will solve the puzzle quickly, and they’ll feel smart because they weren’t given excessive hints. But players who take a longer time will get more information through their interactions with their environments, so they’ll also feel smart because they used that information effectively! Either outcome is much better than dumbing down puzzles or deciding what hints to give in advance.

4:30pm: But What I Really Want to Do is Make Games. I went to the panel because there’s some amount of truth in the title for me; I’m not sure what I expected to get out of it, though, and indeed I didn’t get much out of it.

And in just under an hour I’ll be having dinner with a bunch of interesting people; very much looking forward to that!

Post Revisions: