Thursday’s talks were rather meh, but Friday started off with my favorite of this year’s postmortems and just got better from there; a fabulous way to end the conference. My notes:

10:00am: George Fan, How I Got My Mom to Play through Plants vs. Zombies

The tutorial was the most important factor in helping his mom, so he’s going to focus on that. A player isn’t going to enjoy minute 30 of your game if they can’t get past the first five. So:

10 Tips for Making Your Tutorials Better:

1) Blend the tutorial into the game

Separately labeled tutorials look unfun. Learning in games is actually fun, but people react badly if it’s called out. (Also: separate tutorials often aren’t as fun!)

2) Better to have the player do than read

Try things out in a safe environment. E.g. the one-lane, one-plant-type, one-zombie-type level that the game starts off with. From that, you learn that zombies move right to left, that peas kill them, and about how many peas it will take.

Or: introducing the shovel. Don’t just tell. The place where they were planning to introduce the shovel was also a place where they were planning to stick in a minigame; they thought about a weed-clearing minigame, but that misses the point of using the shovel to swap out useful plants. Tried with wall-nuts instead of weeds, but wasn’t much fun, repeated digging was problematic. They eventually had a bowling minigame where shovels were irrelevant, but was fun, and had the player use the shovel clear out pea-shooters at the start. (Editorial: I’m not convinced that was a great idea, because it doesn’t get at why you’d use the shovel, though I don’t have a better suggestion.)

3) Spread out the teaching of game mechanics

The shovel shows up 5 levels in; money shows up 10 levels in; there’s only one thing you can purchase until the 25th level. The Zen Garden shows up 45 levels into the game, 5 levels before the game is over! That last example is a three-minute tutorial focused on an optional mode of the game; you need to get people invested before presenting them with that sort of thing.

Let players play with their toys before introducing new ones. They were originally shooting for one new zombie every level, but it worked better alternating levels, with new zombies showing up on easier levels.

In-game stores can teach: e.g. put advanced concepts like power-ups there.

4) Just get the player to do it once

Sometimes, a single arrow or flashing button is all a player needs.

Sunflower dilemma: they’re the backbone of the game, but not everybody understands economy. They sound frivolous, less important than getting in defenses. Maybe people would plant peashooters and wall-nuts instead, and would even have initial success with that. The target audience has never heard of RTS games.

Suggestion 1: make it more like traditional tower defense, giving resources from killing enemies. But that removes a differentiator and an iconic character.

Suggestion 2: add more tutorial messages: “sun is like fuel”. They did a little of this, but it alone isn’t great.

Suggestion 3: start with a column of sunflowers. Gave an indication of how many people would need, but would players adapt when the game stopped doing that?

Suggestion 4: reserve spaces for sunflowers. Decent idea, but adds complexity.

Their eventual solution: before, sunflowers and peashooters cost 100 sun, you start with 200, and have choices as to how to spend it. A player with RTS experience will buy two sunflowers; a novice player, in contrast, will plant two peashooters instead, which is exactly the wrong choice. So they changed the numbers: sunflower costs 50, and you start level with 50. This means that almost anybody would buy a sunflower at the start. And a sunflower would be the only choice more often in the game, so that would reinforce the tendency to buy it.

This helps guide the novice, but doesn’t feel like “easy mode” for experienced player. But it alone isn’t good enough: wall-nuts cost 50 as well. They tried bumping wall-nuts up to 75, but it didn’t feel right. Solution was to add an initial charging period for wall-nuts (and potato mines), so you couldn’t buy them at the start.

This was great; the down side was that he had to rebalance the entire game because of this.

5) Use fewer words

Goal is max 8 words on the screen at any given time. And use as few sentences as possible. Tell people what to do, don’t give lengthy details and explanations. Think of it as “the sophisticated caveman”.

Break it up into small chunks, clicking through one at a time.

Crazy Dave was initially intended as a tutorial character. But if he shows up in level 1, he needs to introduce himself, and speak in character. That makes the tutorial less direct. So they delayed his introduction until level 5.

6) Use unobtrusive messaging if possible

Don’t break flow; don’t use popups unless absolutely necessary. Put info on the screen, but let people keep on playing while reading it.

7) Use adaptive messaging

There’s a message that comes up suggesting that people plant peashooters further to the left; you only see that if you plant your peashooters too far to the right, so players who don’t make the mistake don’t feel talked down to.

Another example: in early levels, if people have fewer than three sunflowers a minute into the level, they get encouraged to plant more.

But: leave room for exploration, don’t handhold all of the strategic ideas. Example from a fish game: a carnivore fish eats small guppies; you get hints the first two times it dies, and get told exactly what to do the third time it dies. People who figure it out earlier feel good.

8) Don’t create noise

Don’t cry wolf: only put words in front of players if they’re required info or entertaining.

9) Use visuals to teach

Rule #1: you should be able to look at a plant/zombie and know what it does instantly.

E.g. three-peater.

Rule #2: if you can’t achieve rule #1, it should be clear after seeing it in action once.

E.g. jalapeƱo pepper, pole vaulter. Repeater does double damage; experimented with alternate ammo (shoot swords!), but eventually hit upon: send twice the peas, and send them in a burst so you can see the doubling in a single glance. Or puff shrooms: wanted something less effective, communicated that by limiting the range of its attack. (As opposed to trying to design a projectile that looked weaker by an understandable amount.)

10) Leverage what people already know

Plants chosen to get stationary towers (it was in the tower defense genre) while allowing room to inject personality. Zombies chosen because they move slowly and because they’re bad and you don’t want them to get into your house.

Other examples: coffee beans to wake up nighttime plants; normal zombies \< zombies with cones < zombies with meal buckets; sunshine to grow plants. Plant names are purposely descriptive, e.g. peashooter, squash, consistent -shroom suffix for nocturnal plants.

Teach the player so well that your help section in your menu can be made into a joke.

11:30am: BURN THIS MOTHERFATHER! Game Dev Parents Rant

Graham Devine

Veterans have laid waste to the industry. Only kinds of large games: FPS, RPG, RTS, anything else is too risky. Indie devs get a risky first game published, but steered towards traditional areas for seconds games. Mobile space full of freemium, giving away stuff sucks. Bring back the 80s!

Toby Saulnier

As a parent, I rant a lot. “Get off the computer! Stop playing Minecraft!” Once she thought it as cool that he understood games, gave her powerful metaphors to use. But it doesn’t work, her son doesn’t do his homework.

Monetization problem? How much does she have to pay him to work in the real world instead of a virtual world? Misses the baby boomer work ethic.

Jason Della Rocca

No training wheels for games: e.g. racing games are really hard to pick up. Our kids don’t start with Pong the way we did. How to learn game literacy? Do so in a nutritious way?

Perrin Kaplan

Role playing as woman graduating in 2020 with a degree in video games. Just a dream; schools aren’t going to do that, people in the industry need to help nurture the next generation. Kids always dream, including dream about making video games. We have an obligation to teach, mentor, guide.

Chris Hecker

Duct tape award goes to Scott Jon Siegel. Who laments his overtalking and underacting in his acceptance rant. Don’t tell people to take risks if you aren’t; don’t tell people to make something personal of you aren’t; don’t give many talks and ship few games.

Back to Chris: dysfunctional three-way. Developers, players, press. Players playing/talking about same games over and over again. Press: bizarre focus in previews and reviews, focusing on minutia instead of big picture. Developers: making same games. Appetite for sameness: not just tolerance, actively seek it out. He doesn’t get it to the extent he feels slightly insane.

Dragon speech by Chris Crawford, GDC 1992. Left the industry. Don Quixote.

We are all simply reactive. We all need to be proactive. Players: request and purchase true variety. Press: provide context, hold players and developers accountable. Developers: “we make games we want to play”, so want to play more varied games!

Jade Raymond

Grow up! Stuck in our smelly teenage years. Games have evolved amazingly quickly from a technical point of view, but so many topics are taboo. There are some games that address taboos, but they’re very much the exception. Where are games touching on the Arab Spring, the class divide, internet freedom? All in the news this year, all relevant to games. Circle of life; economic justice; religion.

She’s a studio head; not going to make one of these the core of a $60M new IP. Weave into existing games, though: GTA could comment on penal system, Call of Duty on sexism, Splinter Cell on ethics of interrogation.

Manveer Heir

Stop making broad proclamations about what is/isn’t a video game. There is no one right way.

(Excellent, excellent rant that I won’t begin to do justice to. Don’t be driven by fear.)

Christina Norman

DLC. Can be bad, can be good, players shouldn’t judge in advance.

James Lantz

Against branching conversation systems. Yawn.

Frank Lantz

Ambition, specifically the lack of it. People architect cities around games, e.g. basketball. Video game designers should dream of operating on that same scale. Games living for thousands of years. Too accustomed to thinking of video games as products with shelf lives. Most ambitious targeting Citizen Kane; but that’s inconsequential compared to basketball, go. (Mini-rant re U.S. government shutting down online poker.) Bridge, specifically the development of contract bridge. Chasing after metagame scores, DAUs is embarrassingly modest.

He loves small, personal, idiosyncratic games, too. He’s just speaking to the egotists in the room. How can I make a game that can be seen from space?

Don’t let chess lap us.

2:30pm: Brian Sharp, Concrete Practices to Be a Better Leader: Framing & Intention

My favorite talk of this year’s GDC, I broke it out to its own blog post.

4:00pm: Andy Nealen, Minimal vs Elaborate, Simple vs Complex and the Space Between

New title: What and how I think about game design.

Einstein quote: not everything that can be counted counts, and not everything that counts can be counted.

Likes designing very small experiences, but plays all sorts of things. Question: what does it mean to label any of them as “complex”? None of what follows is scientific, all anecdotal, a hypothesis.

What is complexity? How to measure, where does it start and end? Why would we want to use it?

Some answers to the latter: emergence, surprise, depth. (Another laden term.)

What is complexity? Size of state space? Number of choices per second? Pieces of information that influence these choices? Number of links between elements? Size of state space / decision tree?

There are recurring themes when traversing a state space: e.g. ladders in go. The texture of traversals of playing the state space is its depth. Influenced by players: if nobody plays (nobody traverses), no depth.

Stuff. Things we can count; things we can’t count.

Chess: large state space, even larger decision tree. Go even crazier.

Drop 7: large, but more manageable.

But: we prune the decision tree when playing. We perceive complexity differently than we measure it.

Influencing perceived complexity: creation, reduction, addition. Tools for these three: procedural generation, simplification, coupling.

Texture. Regions that adds stuff without adding additional complexity: different parts of a textured region aren’t fundamentally different. Use procedural generation of texture to generate complexity.

Simplification. Cannabalt: people always want to move right, whereas jumping is irregular; so get rid of controls for the former, only allow control of the latter.

Coupling. Link one entity to another, or interpret a resource differently: doesn’t add entities. E.g. souls in Dark Souls: use to buy a better sword or to level up your stats? Osmos: food -> size -> momentum -> food. Dangerous: coupling is frowned upon in software engineering for good reason.

Part two: design example, Grow21.

Jonathan Blow: “Do not make the player feel smart. Make the player smart.” Andy: make a game that makes him be smart!

Standard deck, split into two shuffled piles, one for each player, divided by color. Lay cards alternately, pulling off groups of one color that add to 21. Only allowed to be one connected component, so if it splits, smaller one gets taken off.

Constraints: knowledge in the world; spatial, no board required; one simple mechanic, few choices; readable; compact; deep.

Solutions to constraints: two player card game, symmetric; adjacent card placement; draw a card and build stable groups; all cards are hidden, no hand; single connected component; set packing is NP complete.

Sarah Elmaleh on Grow21: “I mean, I’m pretty sure I saw the pattern of the universe laid out in front of me last night.”

Procedural generation (by players?), simplification, coupling. Suddenly addicted to physical things, bought 40 board games over last month.

Takeaways: designers can influence and direct perceived complexity. But each version is a different game. (Which is amazing!)

Use generative procedures and leverage texture similarity. Simplify and reduce degrees of freedom. Think of adding links, not entities.

Sol Lewitt: “The idea becomes the machine that creates the form.”

Post Revisions:

This post has not been revised since publication.