GDC 2011

[Update]  I’m back in Seattle and (mostly) recovered.  Thanks to everyone that came to either of my User Research or Game Design talks.  Special thanks to all the readers that introduced themselves; it seems like I am reaching my intended audience of “veteran game designers”.  A narrow demographic, to be sure.  Unfortunately, I spent too much time networking and not enough time attending interesting sessions, but I may cheat and do a wrap-up based on the GDC Vault at some point.


No posts this week.  Between a vacation and delivering two talks I just don’t have time to write anything worthwhile.  If you are going to be in San Francisco, you can catch me here:

  • March 1st at the Game User Research Summit on “Useful Research: Recommendations Designers will Actually Act On”
  • March 3rd at the Game Developer Conference proper on “Design in Detail:  Tuning the Muzzle Velocity of the Plasma Rifle Bolt on Legendary Difficulty across the Halo Franchise”

If you are in the audience and read the blog, please come up afterword and say hello.  I’d love to find out who is reading this thing and get your feedback face-to-face.  I’ll also be posting summaries and responses to the lectures and panels I attend when I get back.

GDC 2010: Design in Detail IV

How do you develop your sense of balance for paper designs? It can be done, you can look at a paper design and have an intuition about how it will work. You are looking for the role, and for a couple key factors that are the results of the role.

The first thing you are looking for is to make sure the paper design is neither too simple, nor too complex. If your paper design is one sentence long, it is probably too simple. People are going to reach the limitations quickly and stop playing. (Remember, balance is longevity)

If it is more than a page long, it is likely too complex. People will never reach their comfort level, they will stop playing because they can’t understand what is happening. Balance is a barely manageable number of choices. (I bet I end up writing a post on this slide at some point.)

This is where roles come in because they provide depth without increasing complexity.

The Sniper Rifle is the best weapon in some situations. The Sniper Rifle has a clear role, times and situations where it is the best. The payoff for using the Sniper Rifle is different depending on what situation you are in. This is good, because game theory tells us that if all the possible strategies have the same payoff, players will pick randomly. As game designers we want to avoid choices that don’t ultimately matter.

Roles are also good because they cause asymmetry, which demand movement. There are incentives to move from one strategy to another depending on the situation.

One of the tropes of the design community… Rock, Paper, Scissors. But it’s a terrible game!!! Every choice has the same payoffs, so you pick randomly. There is no player agency if their choice is meaningless.

It’s a cool shirt, though (

This was easily the most controversial slide in this talk. Several people took issue with the fact that RPS is a bad game because many journalists and designers improperly use it to describe a situation where one unit strictly dominates another, like in a good RTS. But imagine a RTS game where you could only pick one unit, you had to pick it before the game started, and if you picked wrong you couldn’t possibly win, It’d be a bad game!
The reason a RTS game works is because it isn’t RPS:
– You can play mixed strategies (choose more than one kind of unit)
– Strategies have different costs to play (tanks cost more than barbarians)
– You can change strategies mid-game
– Strategies rarely have an all-or-nothing payoff (10 Air units can usually kill 1 Anti-air unit)

So I am not using RPS in the casual sense of “a game with counter-strategies” but as defined in game theory; Hopefully that clears things up a bit.

I also got a lot of people saying, “RPS is the foundation of Street Fighter!” This is true, somewhat, more than the RTS case, anyway. but imagine a turn-based game of SF where the first hit wins the match. Again, a bad game.

The reason SF works is that it is also not really RPS:
-SF is a series of RPS interactions, so things like reputation and anticipation come in.
-It is played very quickly, so low level decision making and muscle memory determine your strategy more than choices, so it isn’t truly random.
-And even with that, most non-expert players tend to “button-mash” which is a great example of “random strategy”

Believe me, I am not trying to insult RTS games or SF (or even Ro Sham Bo Tournament Champions) but to encourage designers to see how roles lead to non-equal payoffs, and therefore avoid random strategies.

“Rock, Paper, Scissors, Lizard, Spock” is even worse game design. (This game is from the show “The Big-Bang Theory”) It looks more interesting, but it isn’t; it is just more complicated. It will still reduce to equal payoffs and random play.

This doesn’t even make sense! (Interesting note: Halo 1 was this close to shipping without a Shotgun, which makes even less sense…)

Avoid strict dominance…Wait, what is strict dominance?

Who picked the Health Potion? The single-use health replenisher you can buy for 30 rupees?
Who picked the Piece of Heart? The totally unique health-extender you can never buy?

Right, everyone picks the Piece of Heart. (If you didn’t, it’s ok, you were probably eight.)

Iterative deletion means you remove all the dominated strategies, then you remove all the strategies that were only good against those strategies. If you cut the Tank, cut the Anti-Tank Mine. Often when you are making cuts in the final stages of production, it seems like a good idea to cut a little bit from a lot of places, but it is uaully better to cut an entire game mechanic and all of the game elements that use it. Otherwise you end up with several systems that feel like they have missing pieces, rather than a single system that is entirely absent.

After GDC I got challenged to come with an example like Rock, Paper, Scissors that is good game design, so I invented Pirates, Ninjas, Sharks. Each has their own strengths. Ninjas are better at night, but Sharks always win in the water, while Pirates come in crews, etc. A good game is one that you can argue about forever.


GDC 2010: Design in Detail III

The first step in Halo 3 was the paper design.

This is one of about 4 slides for Producers.

The Paper Design should happen in Pre-production. You need to do paper designs first so you don’t have 40 artists sitting around while you do it. On the other hand, if you leave it too early, you are going to waste lots of time later. So as producers, you need to find creative ways to give us room in this stage.

Every gameplay object should get a Paper Design. Don’t let Designers hand wave, because we are great at it! If we can’t write it down it means we haven’t figured it out. Lack of design discipline is a huge threat to your project. On the other hand, if you have a designer that has proven his discipline, then trust his paper design. Take off your design hat before you mess up the game!

Nobody has ever seen this outside of Bungie. It’s the original paper design for the Halo 1 Sniper Rifle. It is embarrassingly simple, but it is better to be simple than overly complicated.

Here’s the important things to note:
Some specifics are wrong. That’s because at some point the paper design is abandoned in favor of the existing object in the game.
Some mechanics are unclear. That’s because it is often difficult to capture an entire system inside a single paper design.
Many details are missing. A paper design should only include crucial or unique information, not every trivial detail.
It’s very flawed as a spec, but one thing comes through, the Role of the weapon.

So now let’s look at Halo 3’s paper design. It’s a lot more detailed. But it is still pretty abstract and the role is even more clear. (It ought to be, it was the third iteration!)
It might seem like a waste of time to do a revised paper design since it was essentially the same weapon as the first two games. But it is still important to capture and communicate the vision, because not everyone will have the same understanding of what was important in the previous versions.

“Role: Long-range instant-kill sniper rifle, but reloading makes it difficult to use”
The Role is even more clearly called out. Someone pointed out that Halo has two sniper rifles, which seems to violate the “one weapon per role” rule. Truthfully, they are right, but the missions required enemy snipers using an alien weapon. But, even in that case it is a good idea to give them unique gameplay; that’s why this paper design refers to the fact that the human sniper rifle reloads, instead of overheating.

I felt like this slide failed to capture what is really important about writing a paper design, so I wrote a much more detailed post on the topic.


GDC 2010: Design in Detail II

In order to develop a sense of balance, you need to understand how your brain works. You have an Orbito-Frontal Cortex; It’s called that because it is located behind your eyes, but it’s really your Gut.  When you learn something new, it goes through a process where it builds a model of the world and makes predictions about that model.  If it is right, it releases Dopamine, which cements the model a little bit.

If you are a designer, you need to familiarize yourself with how this process feels, because your ultimate goal should be to get the game inside your head.  You want the model in your gut and the game in the world to be the same.  You should be able to predict how the game will play in a given situation before actually picking up the controller.

Ok, back to Halo 2.  We had to patch, even though we didn’t want to.  Patches are risky and expensive.  Luckily we had network bugs, so we were going to have to patch anyway.  (I’m not sure we would have gotten to patch Halo 2 strictly for balance changes; I’m glad we didn’t find out.)

Choosing what to patch was harder.  You want to tweak everything, but you can’t because then testing gets out of hand.  Games that have the ability to easily change the gameplay often over-patch fopr precisely this reason.

Choosing what NOT to patch was hardest.  We didn’t change the Sniper Rifle because it was right below the line of what we could safely re-balance

Which brings me to my final theme:  Make the hard choices.  Balancing is hard because it requires you to do things you don’t want to.  And it is tricky because there are so many ways to confuse or talk yourself out doing it properly.  But the worst thing you can do is leave the decision up to chance because you can’t make a tough call.

Why are these choices hard?  Again, the answer is your brain.  You also have a Pre-Frontal Cortex.  It’s called that because… Who knows?  We call it your brain.

It is a poor tool, but it’s what we have.  You can’t reason out everything, in fact you can’t reason out very much at all.  There are so many ways that your logical mind has to trick you.  You must confine yourself to reason on the detail scale.  You just can’t hold enough factual information in your brain to make rational decisions about very complex situations.

Radiolab is a great show on New York Public Radio.  They have a podcast, you should subscribe!  In an episode called “Choice” they describe this experiment where psychologists give people a number to memorize, 2 digits to 10 digits.  Then they send them to another room to repeat their number.  On the way, they have someone interrupt them (All good psychological tests are about fooling the subjects) and ask them if they want an Apple or some Cake.

The people with short numbers pick Apples at a high rate; Apples are better for you, fewer calories, watch your waistline.  Those trying to remember longer numbers more often choose Cake.  They are so busy with numbers, they make the decision emotionally.

That’s right, 7-10 numbers are enough to completely fill your rational brain!  My high school calculator had more horsepower than that!

So when you have to think rationally, think about details or you will get hopelessly lost.  And fat.

Ok, here are the four themes of my talk:
– Balance is longevity
– Balance in passes
– Develop your sense of balance
– Make the hard choices

I am going to use these themes to explore the Detail.  Now let’s get to the Sniper Rifle!


GDC 2011 – Design in Detail: The Sequel

Warning: IdeasI’m going to be in San Francisco for GDC where I’ll be giving a talk called Design in Detail: Tuning the Muzzle Velocity of the Plasma Rifle Bolt on Legendary Difficulty Across the Halo Franchise.  It is a follow-up to last year’s talk,which EDGE called “one of the most fascinating sessions of GDC“.  The slides for that talk are publicaly available, but I thought I would lead up to this year’s talk by putting up a director’s cut of the original.  (It will also give everyone something to read while I am on vacation.)

GDC 2010: Design in Detail

Ok, so when I gave this talk at GDC I didn’t have time to finish, even though I was talking very fast.  I did better when I gave it at Blizzard, until my voice ran out and I could barely speak.  So I probably should have cut some ideas out and saved them for this year.

I do see the irony of failing to trim a talk that advices designers to drastically reduce the scope of their game as early as possible.  But to me, the value of GDC has always been that it makes me think about my own problems from a different perspective and inspires new ideas as I listen.  So I wanted to cram as much idea-fodder into the talk as possible, not produce a fluid, polished experience.

I did not make the call on the Pistol, I didn’t even know about that call, it was made directly to the cache file after everything was supposed to be locked down, in fact… but I’m not bitter about it.  I did decide to give the BR some spread, which has been the topic of much debate.  Luckily Reach has been even more controversial, so nobody remembers it anymore.  It’s interesting that every single Halo game has had major controversy over the MP starting weapon, I bet you could write an interesting talk about that.

In 2009 I was at the Art Institute of Chicago visiting my family.  (I grew up in and around Chicago.)  I saw one of the most famous paintings of all time and took a picture of it with my phone.  I actually did have the idea for this talk at exactly that moment, at least vaguely.  Any guesses as to the painting?

It’s “A Sunday Afternoon on the Island of La Grhand Zhot” by Zhorzh Soo-rah (I really wanted to pronounce his name right…heh)  This painting inspired my talk (and its very long title)  Still don’t recognize it?  Here it is from a little farther back.

(I showed you the part in the lower right by the monkey)
This painting isn’t famous for how it looks, or what it shows, but how it was made.

Seurat lived in the 1800’s.  He was very interested in how we perceive color.  Scientists were just discovering that what we see as one color is actually a mixture of different colored light.  To demonstrate this fact, he invented Pointillism, the artistic process of using tiny dots of basic colors to produce an image.  The same way your printer does.

There is a really great play called Sunday in the Park with George that I recommend you see if you are interested in learning more.  You can skip the final act, though.

So I started thinking…  What happens if we take Halo 3 and break it into it’s tiny details…

And analyze just one of them.  Take one tiny Decision, and explore it exhaustively.  Specifically, the time between shots for the sniper rifle.  (I had a lot of fun putting together the pictures for this slide deck.  I never tried to pointilize a screenshot before.)

So here is the actual title of my talk.

If you have been to GDC, you know that most of the talks are too broad to be really interesting.  It doesn’t get much more specific than this!  (Of course, I took this narrow topic and used it to explore lots of subjects and theories, but still…)

My opinions may not represent Bungie management.  In fact, I know they don’t.  (If I only knew the irony of that statement at the time!)  My opinions may not represent reality; this is the past as I remember it, but as we will find out, brains are not reliable.  All examples, even negative ones, are from good games.  I tried to pick on games that everyone knows are great, to avoid controversy.

This talk was really hard to write! Really hard. The more specific the topic, the more there is to say.  It ballooned to almost 300 slides, editing took forever (and it turns out I still didn’t edit it down enough!)

First, some context from Halo 2.

Halo 2 wasn’t just popular, it was popular for years and the Sniper Rifle remained balanced the whole time.  It never needed to be changed, limited, banned, it was still fun.

This can give us a practical definition of Balance:  Balance is Longevity.

(X= 3, I told them my talk would appeal to Engineers!)

Balancing an equation is a process.  But game balance is a state that either exists or it doesn’t.

The 4th rule of Jenga makes it clear (say it with me):

4. Your turn ends 10 seconds after you stack your block

It isn’t balanced unless it lasts.  (I bet you didn’t expect to see a Jenga reference!)

There have been other talks about Halo 2’s crunch; I’m not going to re-hash them.  My really quick personal post-mortem:  Don’t set yourself up to try and fix bugs in the Tutorial and balance the Weapons at the same time, It’s not going to work.  At least not without wrecking your personal life.

Production is always worried about a repeat.  (As you can tell during this talk, Production and I have a love-hate thing going.  It is mostly a joke, but Deisgn and Production have very different goals and the game is only going to work if they cooperate.)  They were always asking me when I am going to be done, so I invented the Balancer’s Paradox.

Balancer’s Paradox
– I can’t balance the Sniper Rifle damage until we set the Player’s health
– I can’t balance the player’s health until we know the engagement distance
– I can’t balance the engagement distance until we set the Sniper Rifle damage.

But after using this for awhile, I actually had to invent a solution…  Balance in passes.  The Sniper Rifle always has to be balanced! You could ship at any time!  So you keep it balanced all the time.  But there’s balanced, and then there’s balanced…

At the end of each pass, the game is balanced to a certain level.  Once a game is balanced to that level, do not backtrack.  Basically, what this means is that if you have balanced the strength of a game element, don’t make it weaker in order to limit the roles of that element.  Unless you have to, of course, in whic case it will impact the schedule so you need to let Production know.

Halo 3’s Balance Passes are listed here.  I think they would work for most games.  They match with stages of game development and I’ll go over them in more detail later.

Remember: Balance is Longevity.  And we could tell by playing it that it wasn’t going to last.  We all had the same feeling and we didn’t really doubt it.  But how did we know?  We had developed our sense of balance.

In Outliers, Malcolm Gladwell lays out his 10,000 hour rule.  This makes me the world’s only expert in trying to balance the Halo Sniper Rifle.  Sort of a narrow expertise, I admit.

Niels Bohr offers this definition of an expert.  This also makes me the world’s foremost expert in trying to balance the Halo Sniper Rifle (and in Driving too Fast for Conditions.)

But what are time and mistakes? They are just experience.  So why is experience so much more important to being an expert than training, knowledge or talent?  Because of how your brain works.


Reciprocal Difficulty

Push on the wall.  This is not a metaphorical encouragement to seek innovative solutions.  Literally, place your hands against a wall and give it a shove.  Now, assuming that you are not working in a cubicle, you have just experienced what physicists call a normal force.  The wall pushed back against your hands with the exact same force you used on the it.

Purchase Gotham City Mutual's Superman Insurance!

Unless you have superhuman strength... Sorry Superman

One of the prerequisites of a flow experience (as defined by Mihaly Csikszentmihalyi in his seminal work Flow: The Psychology of Optimal Experience) is that the difficulty of the activity roughly match the ability level of the participant.  If it is too easy, they will not become completely absorbed and lose themselves in the activity.  If it is too hard, they will become frustrated and unable to make continual progress.  Balancing between these two extremes is the responsibility of the game designer, but often there is not a single setting that works for every player.

One solution is to allow each player to choose their own difficulty level before starting the game.  Unfortunately, the vast majority of players simply choose the default option in their haste to start playing.  Very few players will admit to needing to play on Easy difficulty, and even good players may be too intimidated to start out on Hard.

Every answer seems insulting

I'm tough like week-old bread

Another solution is to dynamically change the difficulty of the game based on an ongoing evaluation of the player’s skill.  But in practice, this automated system usually destroys the intended pacing of the game.  A well-designed game will have periods of less challenge that lead to a more difficult section, building and relieving tension through the gameplay.  A dynamic difficulty system that is too sensitive will add wild fluctuations on top of these natural curves, obscuring the overall effect and flattening out the tension.  One that is tuned to adjust more slowly will often react to the easier segments by ratcheting up the difficulty at the same time that the designer is increasing it for pacing reasons, resulting in an unintentionally large difficulty spike.  Not to mention that the player’s sense of accomplishment will be undercut if they realize the game is “letting them win” by compensating for their low skill levels or “cheating” by getting harder as they get better.

A better solution is to learn from the wall’s normal force and design mechanics that match the player’s skill with reciprocal difficulty.  (Calling this technique normal difficulty seemed confusing.)  In a game with reciprocal difficulty, the more aggressive the player is, the harder the game gets.  But when the player becomes overwhelmed and stops pushing, the game immediately gets easier.  This allows players of all skill levels to be challenged; a good player will play more aggressively until the game gets hard enough, and a less skillful player will proceed at a slower pace until the game gets easy enough.

Examples of reciprocal difficulty mechanics:

  • Recharging Health – If a player is too aggressive and tries to fight too many enemies, they will die quickly, but as soon as they withdraw from combat their health recharges and they can proceed more carefully.
  • Defensive AI – Enemies that are much more effective when fighting from a defensive position are naturally more difficult when the player is pushing hard against them, but allows a timid player to pick them off slowly without putting pressure on them by chasing them.
  • Optional Objectives – Puzzle games that can be solved simply by any player, but have optional collectibles or rewards for completing a puzzle in fewer moves provide more challenge for better players.
  • High Scores – Any mechanic that encourages players to play for a faster time or to score more points is better able to satisfy a variety of player skill levels.
  • Character Leveling – If a player is having trouble with a difficult section, they can make it easier by earning experience and making their character more powerful.
  • Press Your Luck – The player is allowed to periodicaly save their progress or reduce their difficulty level, like going back to town in Diablo or repairing their aircraft in Crimson Skies, but good players will take this option less frequently.