GDC 2010, Day 1

Welcome to the Game Developer's Conference.

Nicole Lazzaro

I was too exhausted and needed to sleep, sorry Nicole. I'm sure you did great.

Yoshiro Sakamoto (Nintendo)

Sakamoto started with 20 minutes describing in excruciating detail exactly which Metroid and Wario-Ware games he did make and did not make. (Short version: a lot of them.) Also he made some series of detective games that we haven't heard about.

He said Metroid is the only non-niche game we would know him for, and in Japan, Metroid is niche (wait, what?) and so over there he's seen as only making niche games. But he's a quirky guy and he likes that.

Early on, he worked with Iwata (current president of Nintendo) on something, I think he meant Balloon Fight. He showed a picture of him and Iwata where each has a thought bubble. Iwata's has a bunch of equations and techie stuff. Sakamoto's has "3 + 3 = 7?" and like a cartoon cat with an arrow to a lunchbox and a lizard or something. They have different modes of thinking, apparently. Recently, Iwata asked Sakamoto how he is able to make such opposite games. The Metroid games are "serious" while the Wario-ware games are totally silly and funny. Sakamoto suspects that actually Iwata's question isn't "how can you make such opposite games?" but is really "how can you make a game with a serious tone AT ALL?"

Sakamoto said to explain, we should know what influenced him as an artist. Early on, he was very affected by Dario Agento, especially his films Deep Red and Suspiria. These are horror films (I think?), and Sakamto said he was so impressed at how the films had tension and heightened emotions. There was some certain kind of music he thought was unusual, but effective. The rhythm had a "dead" quality to it, I think he said, and the music stops entirely at just the right moments.

He was also influenced by Luc Besson's film Leon: The Professional, John Woo's A Better Tomorrow, and Brian De Palma's Carrie. He's also quick to point out that he is not a movie buff, that he has not watched more movies than the average person, that he has not watched all the films of those directors, and that he doesn't wish he were making movies instead of games. It's just that these particular films showed him tools of the craft.

Specifically, he learned the use of these four techniques: mood, timing, foreshadowing, and contrast. He probably should have talked about these in much more depth as this was really the central point of his entire talk, but I don't think he gave specific examples. Anyway, these are the four ideas that he felt were very important to making horror movies work, to have just the right tension.

Then he talked about comedy. He likes comedy and he likes to laugh but a) he is definitely not a comedian (his words) and b) he actually likes making other people laugh more than he likes to himself. Making games that are silly and funny is his way of achieving this, without being a standup comic. He said those same exact four concepts are what makes comedy work. Mood, timing (especially timing!), foreshadowing, and contrast.

Oh, and he also showed us a crazy, indescribable DS game called Tomodachi Collection. You make Mii's (avatars) of your friends, then the game allows you to put them into a bunch of surreal and completely absurd situations. Some are like love scenes on a beach, one was running away from a *gigantic* rolling head of one of your friends, or doing silly dances with them while wearing even sillier costumes, and so on. Sakamoto certainly has a comic touch. Even I started to wonder how he makes a game with a serious tone.

Anyway, his point is that the reason he can do these opposite things--make a comedy game and a serious game--is that they are not opposite to him. They require the same sort of care and he thinks about many of the same ideas in both.

One last interesting thing he said, but I have to translate it a little for you. He talked about how he spends all this time making sure the timing and mood and all that is right, because that's what will create the right emotional response from the player. He was trying to say that he thought of the player as this nebulous thing out there. Kind of like he makes a work of art, then throws it into some sort of void where, theoretically--some humans will enjoy it. I know exactly what he means because I often have that exact same feeling. I've heard other artists mention this same idea too. They are designing something that people are supposed to enjoy or appreciate, but...who are these people? Sakamoto just does his best then hopes for the best.

BUT, then one day he changed his view. After the release of one of those detective games we don't know about here, a woman who played the game liked it so much that she sent him homemade chocolate candies. He explained that in Japan, this is what women do for men to signal romantic interest. He said he was shocked by this, like he didn't know how to even react. It was the first moment he really felt deep inside him that actual real people enjoy his games. Not just theoretical people. So this praise he got had quite an effect on him, and from then on, he pictured specific people when he makes his games. What will his wife think? What will some little boy he knows think? And so on. Well, I thought it was interesting.

Jaime Griesemer (Bungie)

Griesemer's talk was called Changing the Time Between Shots for the Sniper Rifle From 0.5 to 0.7 Seconds For Halo 3. It was about multiplayer game balance, and he covered many similar ideas as my GDC lecture last year and my writings. He even quoted me in his presentation (about the Yomi Layer 3 idea). Also, unexpectedly, quite a bit of his talk was literally about the sniper rifle. He stared his talk at 1:30pm and he said he had 130 slides. He said he started with 300 slides, and had to cut it down, but maybe he can present the rest of the talk at SniperCon 2011. Apparently this a lot to say about sniper rifles.

Griesemer said that balance is something you can't wait until the end to do, and also something that you can't do until the end. He said if any producer people give him trouble, he can say one of those two things as an explanation. (In my opinion: "When will you be done balancing?" can be answered "Pretty much anytime you want. How good do you want the game to be?") He stressed that programmers need to create PLAYABLE builds very early and focus on fixing any bugs that ruin or prevent gameplay before any other type of bugs or features, even if they are features that Griesemer himself requested. Furthermore, builds must run at at least 30fps. These are the basic requirements for him to even do his job at all.

Before there is a build, he stressed the importance of a "paper design." I found it a strange term because it didn't mean creating a prototype using paper and cards and stuff as is often done (like for Warcraft 3). Instead, he simply means a description of what each weapon or vehicle or whatever is supposed to even be. Like, what are its basic properties and why do we need it. This is not something a programmer could go implement, it's more like a vision statement for each thing. Though he stressed this multiple times, I haven personally never found the need for it. I have a very strong sense of what my vision for each card or character or whatever is, and simply discuss this with other designers or testers or whatever.

Anyway, the next part of that idea is that each weapon needs a "role." The sniper rifle has a very clear role: it's supposed to be the best weapon at long range, it's supposed to kill in one hit if you get a headshot, and it's supposed to have a zoom so you feel like you're really a sniper. Likewise, a shotgun has a specific role, a handgun has a specific role, and so on. He pounded on the point that you should fill every role you need and you should have nothing extra. Like if you added another sniper rifle that was kind of crappier, what is the point of that? First of all, it's strictly dominated by the real sniper rifle, second it clutters up the game, and third, it can't even be tuned to have a really distinct role because it's too close to the real sniper rifle. So cut it. A mentor of his said after you make your stuff, cut half. By doing that, you have that much more time to fix up the stuff you have left. He even said that these extraneous options add "complexity without adding depth" which is word-for-word what I said in my lecture last year. So yeah.

Griesemer described the concept of flow (being challenge but not too much, in the zone) and how important of a concept this is. (Yeah, I would go so far as to say it's required to really understand design.) He briefly mentioned the dopamine rewards your brain gives when you are learning and when you have success on a task. Anyway, he thinks A LOT about what keeps a player in flow and what breaks it. If you do a weird control scheme like trigger button punches but face button shoots a gun, then players will be stuck thinking how weird that is. Even if you put inventory on a d-pad or heaven forbid, anything on the click-stick, he thinks these can break flow. But also things like the rate of fire of the sniper rifle. If you allow a crazy fast rate of fire, it's just silly and stupid and you can't get into your role. He thinks there's a certain rhythm of pulling the trigger on the controller, having the shot fire, hearing the sound effect, seeing the bullet's tracer, and how long you wait until you are allowed to fire again--there is a way to do that where you get really immersed in it. That is what he's searching and searching for. If he finds some part of another game that gets something like that right (like, say, a snipe rifle--remember this secretly is SniperCon 2010) then he plays it endlessly to feel that flow and study it.

He also mentioned that his team sometimes gets annoyed that he spends hours on things in the game that are fine, instead of on things that are broken. For example, he will drive a warhog for hours, feeling how it turns and accelerates and so on when the warthog is fine. (Or, you guessed it, firing a sniper rifle.) He could use that time on some other vehicle that feels totally bad, and his team wonders why he wastes his time on the already-fine things. Griesemer's response is that he doesn't want to "imprint" on the bad stuff. He knows it's bad, he knows it's going to change, he doesn't want to get used to it in any way. On the other hand, he finds it very necessary to fully internalize all the things that work correctly.

This leads us to what is really one of his main points. Balance is done via intuition, not mathematical modeling. If anyone's keeping score, that's the third time this point has been made in a GDC lecture. Rob Pardo (Blizzard), me (Sirlin), and Griesemer (Bungie) have all said that exact point. And while we're on that, all three of us made another somewhat related point, too. Pardo and I both said that you have to be careful of mathy people who want to balance the fun out of everything. Like we could bring the power level down to be boring and make things balanced, but that fails at our goal of making things FUN and also making different weapons/characters/whatever extremely DIFFERENT from each other. Griesemer's version of this same idea was the concept of making everything crazy powerful, and balancing that. Don't make everything really weak, and balance that, it's boring. If everything is overpowered, then nothing is overpowered. He WANTS each weapon to seem overpowered when its fully in its role.

Griesemer then addressed the concept of randomness. So lets say that each weapon has some situation where it's the best weapon. Ok. But how easily can you get into the right situation or avoid the wrong situation? If your game is so chaotic that there is simply no way to predict or control situations, then it's just pointless. It's like playing actual rock, paper, scissors. (As I've written before, you need weighted RPS, preferably with unclear payoffs to start making it interesting.) On the other hand, if you have perfect control over which situation you're in...like if you could guarantee 100% that you can fight at far and lord the sniper rifle over everyone, then that's no good either. Completely predictable means boring gameplay, but completely random means stupid gameplay. He wants a certain non-zero amount of randomness.

Furthermore, he said that any comments on this randomness are completely predictable on any project. Someone will say to make it way more random. Players who are bad like lots of randomness so they have a shot at winning. Other players will say to remove all randomness. If they are super highly skilled, then they want to remove all randomness so that they can win 100%. He believes that it's best not listen to them either and maintain some amount of uncertainty, though it's tricky how much exactly.

He listed a few types of playtesters he sees. One is the Optimizer. (This role played by Alhazard of sirlin.net forums.) The Optimizer plays whatever makes him win, nothing else, and can somehow find the most powerful things very fast. Then there's the Role-Player. (Possibly played by Thelo of sirlin.net.) The Role-Player wants to play a certain style no matter what. If it's strong or weak, they'll play it. He says he likes these players because it means even if sniping is overpowered, there will be non-sniper player-testers simply because they prefer whatever role it is they always go for. He can then see how badly they die, if it's close or not. There's also the Pro-Player (waterd of sirlin.net). The Pro-player has things in common with the Optimizer, but is specifically interested the game as a form of competition and he is the first to request the removal of all randomness. He will use tricks and lesser known stuff to win if it works, though an Optimizer only uses the most broken thing period and nothing else. (You could say I was an Optimizer player in some games, though if you really commit to that, be careful, because you aren't learning much else in the game. Better hope nothing can come along to beat you or you have nothing else.)

He said he observed that Optimizers were using the sniper rifle 100%. Even at close range. They didn't even us a scope, just jump around and sniper shot pile in melee range. Even a body shot did lots of damage and you couldn't really get away in time because the 0.5 seconds between shots was so fast.  He said the sniper rifle was an especially bad offender not JUST because it was overpowered, but because it was going outside its role. It's not supposed to effective at close range, so that is a major problem even beyond the balance.

He considered many possible solutions, and I won't go through them all, but many of them just "didn't feel like a sniper rifle should feel" even if they were balanced. The one change that didn't violate any "feel" requirements was the extra time between shots. In fact, he said it improved the feel. The rhythm between shots while sniping from a distance (with scope zoomed in) felt more like, well, sniping when it was made longer.

Another point he made was about how players claim that he sucks at the game. He says that sucks at the game kind of on purpose. His goal is to build a mental map of the game's balance, and if focused on being a good player, this would be harder. He would specialize in some aspect or style or weapon, then be afraid to change it even if it was needed. More interestingly, he said that by working on the problem of balancing things, he is being challenged and he gets dopamine when things go better. If he were a player, he would get dopamine as he got better and did successful moves in the game, and that this would make it hard to tell which thing his brain was responding to. Do I feel good because I kicked ass that game? Or because this balance is working out really well? He prefers to avoid the confusion.

Maybe I'm one of the few people who can speak to BOTH sides of that. For Street Fighter, I did the opposite of what he said. I absolutely played balance changes at a high level, and if someone thought something was too good I made them try to beat me with it. I used my ability as a player extensively. While I can't play every character in the game at a high level, I can play most of them pretty well, so I'm not even that specialized. BUT, in other games I'm balancing, I take the same stance as Griesemer. I actually intentionally don't play them to a high level, and instead rely on the competence of my play testers. The skill is in knowing when to listen to them, and when to stick to my own gut. I build a mental model of the balance, just as Griesemer says, and I am able to make changes that, over time, seem to please the playtesters quite a bit. So I think both methods are possible.

Rob Pardo (Blizzard)

Pardo's talk was good and solid and had lots of information, though it contained no new ideas for me. I've heard him say most of it before, but that isn't to detract from the talk. Amazingly, he managed to say like a thousand and generate no disagreement from me at all. He said his talk was originally designed as an internal Blizzard talk, to get his employees on the same page, but he decided to do it at GDC instead so that he'd actually finish it.

He said Blizzard puts "gameplay first." An obvious thing right? Apparently not, because gameplay-first is not the approach used at many studios. Some put technology first. Some put story first. Some put art first. Blizzard puts gameplay first. What that means is when you have to make a trade-off, and you practically always do in design, then gameplay wins. He gave an example about story and another about art.

The lore of Warcraft was that only night elf males can be druids. When planning World of Warcraft, Pardo told Metzen that they'd have to change the lore so that females and some other race can be druids too. Metzen (keeper of the lore) said NO WAY. The lore is all important you know. Pardo said "Do you want Druids as a playable race or not? It would be crazy to make one of our only NINE classes a thing that has only one race and that can't be female." Metzen said ok fine, and they changed the lore. Gameplay wins.

The animation for mounting your horse was another one. Some people said "it would be cool" if summoning your horse caused your horse to ride in from the distance and as it approaches, you jump on top if it with a cool animation like in Zelda. This is obviously a stupid idea, but for some reason they actually mocked this up. Then it became even more obvious how bad of an idea it was. When a rogue is stun locking you or whatever, do you really want your horse riding in from the distance for several seconds? No, you don't. The summon horse thing in the real game is just a poof of smoke and you appear instantly on your horse. Does it look dumb? Yes. But in this case, gameplay needs to trump art and animation.

Next he talked about "What's the fantasy?" He means to determine what fantasy experience you want to fulfill for the player, then work toward that. He gave an example where they failed on that, the Paladin. A paladin is a guy who saves people, maybe in some ways like a triage doctor can save patient's lives. Then he showed a screenshot of a raid from the point of view of a paladin healer, and the entire screen is covered with dozens of little health bars and readouts and it looks like the most awful thing ever. This is not the fantasy of being a paladin. "Fail," he said.

Like Griesemer, Pardo said to make everything overpowered. He cautioned against letting math people balance everything into boringness. Make big exciting effects and moves units and so on and have the confidence that you can balance them at that level. He also said "it doesn't cost anything to make something epic." Apparently when they first put in a dragon into World of Warcraft, it was about 5x the size of a normal character. Pardo said "that's a dragon? Seriously?' So the programmers gave him two special spells he could put in his spellbar, one that would grow a target monsters by 10% and one that would shrink it by 10%. He said he must have pressed the grow button at least 20 times then said "now THAT is a dragon." He wonders why they even gave him a shrink button because he has "never pressed it once, ever."

He advocates "concentrated coolness." A good example of this was the taking the heroes in Warcraft 3 and making the classes of World of Warcraft. There were three Orc heroes in Warcraft 3 that might have gotten their own classes in WoW. But Pardo wanted fewer classes rather than more, and for each one to be very full of cool stuff. So they actually combined the best stuff of those three War3 heroes into the warrior class in WoW.

Randomness. Pardo says that about every week, someone will come into his office and say that the random number generator is broken. (It's pretty much a rule of all games that people will always claim the RNG is broken no matter what. It's guaranteed to be complained about. It even happened to magic the gathering and it happened to a poker site who then published the exact method used for the RNG to prove it was fine.) Anyway, death, taxes, and people to wrongly complain about random numbers are all certainties. Why is this? Humans are extremely bad at understanding random things. Flipping heads 10 times in a row makes you think something is wrong. But unlikely things happen. Also if heads is failure, of course you'll remember that. If you got success 10 times in a row with tails, you'd just be happy and forget about it. Also, were you flipping these coins all day? Of course somewhere in there you'll get a streak. Were you doing this alongside thousands of other people doing this at the same time? Do you even realize that is practically certain that SOMEONE out of thousands will experience that streak?

Pardo said that instead of shipping himself inside every box to explain to people that random things are random, he instead changed randomness in their games to match what players expect. Instead of real randomness, they use fake randomness on a lot of things. If you fail at something (for example, the monster doesn't drop a bear paw or whatever) then the chance of success increases a bit, until it's eventually at 100%. If you just play through it, it feels random because sometimes you get the drop and sometimes you don't, but you don't get those bad streaks.

Pardo also advised us to turn punishments into rewards. He that people like rewards more than punishments (ha), and yet designers always tend to pull out the stick first. "What happens when you DIE" and so on. An example here was the rest system in World of Warcraft. He put in the system to discourage people from playing these marathon 16 hour sessions, hoping instead to get people to play just 2 or 3 hours each day or something. So after X hours of play, you would be "tired" and your character only earned XP at 50% of the normal rate. Seemed logical as a way to get people to stop playing and have their characters rest until tomorrow. But he said it was universally hated. The beta testers "cried bloody murder" at this horrible thing Blizzard was inflicting on them. They should take it out or die, and so on. Pardo said fine, he'll give you a "reward" instead. So he doubled the amount of XP it takes to level up, then let you earn XP at 200% of the normal rate! But after X hours, you only earn at 100% of the normal rate. And now everyone LOVED it! Even though it's EXACTLY the same system, functionally. People like rewards even if they are fake.

Multiplayer first. This should be obvious to anyone who's thought about it, but so many companies seem to do this wrong. Multiplayer isn't something you "just add at the end." It's pretty crazy to spend like 2 years on single player stuff then add multiplayer. How would that even work? Probably lots of stuff won't even work at all in multplayer, then what are you going to do? Change it to work in multiplayer, then retrofit all the single player? Or another way of looking at it, how long is a single player game? 5 hours? 10? 70 hours if it's final fantasy or something? Ok, that's nothing compared to how long multplayer needs to last. It has to last like thousands of hours. So isn't it pretty obvious that the thing that has to be interesting for thousands of hours needs a lot of dev time? Once you have multiplayer nailed and solid, then you can make the single player stuff. Multiplayer has strict requirements about how fast a unit moves or how quickly it shoots and so on, but it can kind of be whatever in single player. Lock it down in multiplayer, then build your single player missions around that. Blizzard does their single player stuff in the last few months of the project only.

He said a lot more stuff, but that's all you get now.

Chris Hecker

It seems that Chris Hecker spoke on the topic I felt destined to speak on. Maybe I could do it next year and go a step further than him. His topic was "Are Achievements Harmful?" Afterwards there was a question and answer thing, and I asked him "Why do you make a point of putting a question mark there? Why are you not convinced? Why do you present all this evidence yet not take a stronger stance?" I personally feel Hecker is requiring an unreasonable amount of research before shaming the bad guys here. Anyway, I'm fairly well versed in his topic and he presented lots of great information, so let's get to that now.

If you reward people at something, they do better at it right? Well, not really. There are several books he pointed to and tons of studies (more than a hundred) that debunk this. He started by talking about a couple books that take a very strong stance against the concept of external rewards, but he was not satisfied with these books. He thought they had agendas or something, so he went to the journal papers themselves to see what's up. Before we get to that, let's talk about the books' claims.

They are all about how giving people external rewards is the worst thing to do ever, in basically any situation or circumstance you can think of. Get some test subjects to solve some puzzles and test the differences if you pay them or don't. Paying them makes them perform worse, be slower, less creative, have less fun, and so on. Give people gold stars reading a book, and they read books for gold stars instead of the actual joy of reading a book. Stop giving gold stars, they stop reading books. Pay a kid to clean his room, now it's overforever. He'll never clean his room unless you pay.

There was a joke in there somewhere about how giving people a pizza for reading a book would be a bad idea because the effect is that you don't care about reading the book at all and you just do it (resentfully usually) to get an (unhealthy) pizza. But if you gave people a book for eating pizza, it would work out better. Then people would like to read books and not be so hot on unhealthy pizza.

I'm really skimping here and not doing this justice. The range of examples where giving external rewards hurts is staggeringly large. Performance reviews, bonuses, bribes, candy, anything you can think of. Forgive me for doing a bad job summarizing this part, I'm exhausted.

Ok, then Hecker questions this premise. Not like "I think it's wrong" but "is this for real or what?" He looks at source papers and finds that these papers all argue with each other like kids flaming each other on internet forums. They disagree deeply and have conflicting findings. But notably, if you look where it "mostly" ends up, the research does lean toward the side of "external rewards really are bad and do the opposite of what you'd want." It's just that there's some amount of error +/- that surrounds these things, and it's close enough to the center that you could make an argument the other way, even though it leans toward supporting the above claim. Well...ok. Though Hecker points out that in all the debate, there are a couple things that both sides to agree on. These two things are enough to sound the alarms in my opinion, and I still wonder why Hecker doesn't. To understand those two things, let's first look at the components we're dealing with.

These studies break things down into dozens of categories. What if the reward is tangible? What if it's verbal? Or symbolic? What if the reward was expected vs unexpected? What if the task was boring vs interesting? What if the reward was endogenous vs exogenous (as in, read a book to win a book vs read a book to win a donkey)? What if you're given just information ("you killed 5 orcs") vs controlling/suggestive ("you killed 5 orcs, as you were supposed to do"). There were a whole bunch more, you get the idea.

Ok, so there are studies on every which way you can think of combining those things. Some combinations don't have much effect. Some have lots. There's a couple that BOTH sides of the aisle in psychology agree about though. I'm too tired to remember the exact combination of variables that were the two most important batches, maybe I can edit this post once someone tells me the exact sets. But I do remember the gist of it, which goes like this:

1) If you have an interesting task, and you give people rewards for it (like gold stars or achievements in a game), then people have much lower motivation to do that interesting task anymore. They do it ONLY for the bribe.

2) If you have a boring task, that same kind of bribe doesn't make people want to do the task less. They already didn't want to do it because it's boring. But instead of not doing it all, now they will do it for the reward. So for boring tasks, the bribes really do work.

In other words, you can basically pay people to play games that are boring. Hecker points out that we are sucking pretty bad if that's what we're doing. I mean aren't games supposed to be fun to play without bribes and stuff? He half-joked that it's ironic people would make games that are boring and trick people into playing them so that the game makers can make a lot of money, when money itself doesn't lead to happiness if you're above the poverty level (supported by like a thousand studies, btw). He also outright said he has pity for any developer who spends his life making a game that is boring and dressed up with external rewards to bait people into wasting their time on it. Maybe make something that somehow enriches people's lives instead? Or that makes an artistic statement? Or that has some sort of positive effect or fun in it?

Hecker made one point that I have never even thought of. He talked about how these days we want to do more and more metrics. Like what if we make the hat a ligher shade of pink, can we get more people to click on it? What if the slot machine spins a bit faster, and so on. The thing is, what is easy to measure and what isn't? What's easiest to measure is when you change the external rewards and test if that makes people play the game more. But think about that for a minute. We just saw from all those studies that external rewards have the biggest impact when the game was boring to begin with. So by obsessing about measuring everything, you will tend to design things that are capable of being measured. In short, you will design boring games that need external rewards in the first place.

I guess I think this whole subject is a lot more dangerous than Hecker thinks it is. Having an entire industry that makes people into zombies who play boring things over and over because their brains are unable to resist the psychological tricks that external rewards use...just sounds terrifying. And also a huge waste of time for everyone involved. It's also dangerous to erode away a generation's grounding in the importance of internal rewards. Those who are internally motivated try harder, do better, have more persistence, have more fun, etc, etc. Not just in games but in life in general. So an industry that kind of brainwashes everyone into this external rewards mindset is dangerous in ways beyond just the time wasted. Once you become conditioned to behave according to external rewards, you'll find it that much harder to study for studying's sake, to improve for improvement's sake, or even to enjoy the sunshine in the park for the sake of enjoying the sunshine in the park. If you don't unlock an achievement for enjoying life, I guess it's not worth it, right?