The Truth in Game Design
Jonathan Blow gave a talk recently at Champlain College. The subject of the talk was a game design philosophy in which you ask questions of the universe and then are open and attentive to the answers. In Jon’s view, game systems are like scientific instruments that can reveal complex and fascinating truths about the world. It’s a terrific talk, and you can download the whole thing here.
Jon’s talk made me think of what a couple of brilliant game designers said at GDC.
In two separate talks Sid Meier and Rob Pardo talked about how players consistently misunderstood, and were frustrated by, randomness. For example players, like most people, tend to subscribe to the “gambler’s fallacy”, the mistaken intuition that random events are spread evenly over time instead of clumping, well, randomly as they actually do. This is the fallacy that leads people to expect a flipped coin to be less likely to land heads after a run of heads.
Pardo and Meier both described the same solution to this problem, which was to alter the behavior of the game to correspond more closely to the player’s intuition about how randomness should behave. For example, if you have an event that is 50% likely to occur, and it doesn’t occur, then you make it 60% likely on the next attempt, 70% likely after that, and so on until it is certain. Voilà! A coin that is less likely to come up heads twice in a row, and never has a run of more than 6 heads.
Now, Rob Pardo and Sid Meier are amazing game designers and it’s a privilege to hear them speak. Both talks were full of invaluable insights from practitioners at the top of their profession. But this particular detail really stuck in my head, and rolled around there with Margaret Robertson’s microtalk about behavioral economics and Chris Hecker’s talk about external reward systems, and the numerous discussions of Zynga’s quantitative, behaviorist, social-game design methodology that loomed over the whole conference like Chernabog glaring down from Bald Mountain.
Eventually this tiny detail, this thoughtful little adjustment of the pillow beneath the player’s head, became emblematic of something big and important at the heart of game design:
Shouldn’t games be an opportunity for players to wrap their heads around counter-intuitive truths? Shouldn’t games make us smarter about how randomness works instead of reinforcing our fallacious beliefs? Shouldn’t games increase our literacy about interactive systems and non-linear possibility spaces? Isn’t contemplating the elusive truth about these things one of the most powerful cognitive benefits of a life spent gaming? And isn’t it therefore our job and our responsibility to guide players along that rocky path, no matter how uncomfortable it might seem, and at least give them a chance to glimpse the truth and begin to approach it and acquire something far more valuable than comfort, rather than making them a bed on the near side of that stony ground?
Ok, so maybe not all games have to confront players with difficult universal truths, some games should indulge our superstitions and muffle us in comfort. But we’re talking about SID MEIER and ROB PARDO here! We’re talking about Civilization and Starcraft! We’re talking pinnacle of computer games, games whose legacies will reverberate for generations. And if we have used computers to build intentional flaws into the numerical heart of our deepest and most cerebral games instead of using them to elevate our understanding of the computational heart of the universe, then we’re doing something wrong.
One of the greatest gaming experiences of my life was the way that Poker forced me to re-program my own kludgy understanding of probability. It took me a few years of dedicated play and study, years that included as much suffering as they did joy, and moreover in which the joys were somehow uniquely bound up in the suffering. But eventually I got to a place where sometimes I get a little glimpse of the world-as-probability, of the Bayseian logic that boils beneath the solid cause-and-effect surfaces of my everyday, results-oriented consciousness.
And let’s not forget that Poker, which refuses to bend to our cognitive biases, which takes our naive misconceptions about the world and beats us mercilessly with them and mocks our numerical ignorance and offers only the most dedicated, hardworking players an opportunity to slowly and painfully approach the truth, also happens to be more popular than Farmville.
I saw the same two talks at GDC, as well as several others explicitly about player deception, and I interpreted them very differently. The goal of the kind of deception advocated by Sid and Rob is to help construct a shared fantasy that can provide real experiences without being based on real world truth.
On the other hand, the kind of deception inherent to business models like FarmVille is not in the service of creating new experiences and is instead out to maximize short term profit.
But coin flips and computer-generated random numbers are actually really *weird*, when you think about it—few random processes in nature so thoroughly lack any kind of inertia. That’s why they’re unintuitive: the outcome of many (if not most) random processes we observe in the world in some way depend on the previous state of the world. (Poker is a good example of this—if you see a jack of clubs on the table, you know you’re not going to get another jack of clubs dealt to you.)
For me, one takeaway from this whole discussion is that it’s that there’s nothing “neutral” about plain-jane uniformly distributed random numbers that you get from a d6 or a call to random(). It’s worth thinking about how randomness works in a game design, and that using other models for randomness (normal distribution, Brownian motion, Markov chains, etc., where the previous state of the game is taken into account) can have a huge effect on how the game works, sometimes for the better.
SID MEIER and ROB PARDO are screwing with the laws of nature.
However, how do you have high scores that are meaningful in a game like, say, Drop7, if there is the possibility that one gets a random run of the most awesome non-1/2 discs? That high score will tower above others, and will have more to do with persistence than with skill.
I think it makes sense then, that in a game like Drop7, only average scores should be compared. However, that game publishes high scores to Facebook.
Is Sid Meier actually making games though? As you remove randomness you drive the experience more toward a scripted experience. Couple that with the level up over time mechanic in most of his games, they become a flat OCD-driven experience that doesn’t really feel much like a game, more of a check the boxes exercises. Or maybe more like a toy that has certain limits and constraints.
Perhaps a good compromise is to make the “difficulty” of the game attached to the randomization model employed by the game. That way the player knows that overcoming the challenge of the game occurs when they can adapt to the true probabilistic spread of the game system.
Sort of like iTunes and how they randomize the song playlist to have more of the same artist play back to back, or less together. The trick, I guess is, is to mask this so that the player doesn’t feel patronized; this I think is the most difficult task, but one that points more directly to the psychologically irrational beliefs of the player.
I like to think that I wont design intelligence that forgoes intelligence, but then again, I’m not designing games that will be played by millions on day 1…
The problem isn’t that people can’t understand randomness, it’s that predicting the results if a die roll isn’t a very good game mechanic. It’s not fun to work for twenty minutes to stack the odds in your favor and still lose because the ai rolled a natural twenty. The result of the actions that both Sid and Rob described isn’t a game that feels more authentically random, it’s a game that is no longer actually random, and predictable mechanics are easier to make plans about, making them more fun. What they ought to do (IMHO) is stop using die rolls as a fundamental mechanic completely.
Great points Frank. Here’s an interesting example of how I cheated randomness in a game: I love Catan and play the XBox version a lot as a single player vs. the computer. There’s an option on the game to tweak the randomness of dice rolls so that while the rolls are still random in order, after 36 rolls the true double dice pyramid is achieved. In other words, you won’t see a run of 15 6’s before a 8 ever comes up.
This solves and breaks Catan at the same time. When expert players face off in Catan, it becomes craps and the game loses its lust. So this helps fix that while still supplying randomness and keeps Catan fun for me. But at the same time this control over randomness ruins the true essence of Catan and fails the natural laws of game design. So even as a player I find myself caught in a paradox.
Adam, it’s true that “pure” randomness doesn’t occur much in nature. And most cognitive biases are less irrational than they seem when you consider the world they evolved to process. But the larger point remains, we modify our games to make our players more comfortable. And this is a great impulse! But we should be aware of what we’re doing, and of what we potentially give up by doing it.
Minimalist, I’m pretty certain that the highest scores in Drop7 come from players who have built up deep heuristics and analytical skills in the game, and are able to *take advantage* of the lucky opportunities the game gives them. I think some of these players are on a spiky journey that goes up and down but trends up over time. Their highest peaks keep getting higher and that’s one of the things that keeps them going.
IOW, QQ more, noob! :)
Keyvan, “The trick, I guess is, is to mask this so that the player doesn’t feel patronized…” I think we mask too much and patronize too often. I would like to encourage game designers to stop thinking of players as subjects of psychological experiments and think of them as collaborators, fellow researchers in the experiments games allow us to do on ourselves.
Charley, wouldn’t truly serious competitive play using the “true pyramid” require tedious bookkeeping? Keeping track of all prior rolls becomes a necessary pain in the ass.
Isn’t this one of the puzzles of information theory?
If uncompressability is a measure of information–which is a pretty common definition–then random noise happens to carry the most information.
But that doesn’t make sense to our usual manner of thinking. How can random fuzz be considered informationally dense? Mathematically, static might have a high information coefficient. But to me, it’s just a missing channel on my TV
So, it must be patterns, then, that we think of as information. And if that is the case, then rigged random number generators in games create a new pattern to discover. And being predictable, patterns allow us to master them and move on to higher order thinking.
Maybe our wizened old game developers are pulling the wool over our eyes, or maybe they are just trying to create order out of chaos.
I agree whole-heartedly with the argument. (In fact I agree so much that I wrote an article on the same subject with the same title for Gamasutra last month! http://www.gamasutra.com/view/feature/4263/truth_in_game_design.php)
That said, as much as these design choices make the designer in me cringe, I could see the benefit of obscuring a detail like this if it helped put a focus on a higher-level truth somewhere else in the game. As a top-of-the-head example, it might be difficult for civ players to pay attention to the nuances of negotiation during the end game of civ if they must also continue to worry about micro details like losing a major city because of successive bad rolls to a much weaker unit. If doing so helps more players walk away from the experience with a useful metaphor for how to deal with the real world, then that’s a good thing. Harm will only come when you tamper with the truth so much that you convey something false, or worse, you end up offering nothing at all.
Great post!
The other night I had the foundations of my reality scrambled by poker. This isn’t the first time this game has done this to me. Your article described many of the troubles my brain was facing. Games are awesome windows to truth. They are also awesome lenses that filter truth and provoke questions. I was stimulated and confused on my way home from losing my $60. (thank god I got out of there, I smelled a losing streak coming!)
“…the way that Poker forced me to re-program my own kludgy understanding of probability.”
I really love and enjoy most artgames, with their messages and sincerity, but this sort of personal growth is just so practical and almost without question more powerful.
Please write more Frank. The world could be yours!
Thank you Mike. I think that maybe it already is!
I guess this is why seasoned poker vets despise any game with a wild card…
This happens all the time in online backgammon. People think that the dice algorithm isn’t fair or isn’t truly random. The problem is that people remember what they perceive as unusual. No matter how unlikely something is, like getting 10 pairs of sixes in a row, given enough dice rolls, it will eventually happen to someone. And you can bet this person will use that as evidence that the dice algorithm is out of whack.
As a side note, I learned from John Conway that in Backgammon tournaments they roll all the dice beforehand so they can give everybody the same rolls. He says the universe might be like this and you could never prove it. It’s obvious to John Conway that the universe isn’t like this. But it’s also obvious to him that you could never prove it.
http://www.flickr.com/photos/franklantz/3482233300/in/photostream/
I think the truth most players are after is about how well they are playing. It can be difficult to distinguish that from the noise of random dice. Poker players have to learn the truth about randomness at the same time as learning about how they are doing, but they have to play thousands of hands to do so, and most computer gamers would like to know quicker than that.
After initially being strongly against this, my position is therefore now that adding in some correction to reduce the randomness of the average random effects on a player is actually a decent idea.
The truth of the matter is randomness is over-rated, there can be degree’s of randomness but they should lean more towards a designed kind of “randomness” not the real kind.
Many of the best games come from static and predictable designs which confine the player and his actions to a particular space (think competitive deathmatch levels with good ‘flow’). Why is it that some deathmatch levels are played more then others? This is because players aren’t really looking for randomness they are looking for the level to constrain their freedom to some extent and to limit their options from infinite amount of actions to a finite amount of meaningful choices.
Games are ultimately about making _meaningful choices_ once you give the player everything and have “all possibilities” on the table you actually create situations in where no choice is meaningful.
The great thing about older deathmatch games was their constraints and their arena like nature to keep the game moving. This is a kind of hard determinism that is absolutely necessary for the game to be any fun at all.
I don’t believe in “game design in theory” I believe you have to test your designs and protype them in real time to get meaningful feedback, there are too many “gotcha’s” about doing game design from your head alone.
I think this is why civilization’s sequels have been so conservative about changing the design of civ, they were scared they would break the game in some fundamental way.
Game designers often have no clue what makes their games fun, if they did they would not make the same mistakes they have been making for 20+ years, there is much to be learned from the conflict between players feedback and the designers themselves.
that’s crazy talk, there is no “designed randomness” it is either random or it ain’t
This reminds me of my experience playing the original Dragon Warrior on the NES. I’d played RPGs a lot as a kid, but had since gotten bored and even embarrassed by them, but the Dragon Quest series re-sparked my interest in the genre because of randomness. The “random” battles in a game like, say… Final Fantasy VII are hardly random at all. They come in regular-ish intervals. You NEVER have a battle immediately after another one, and you never can run too far without a battle, either. It almost becomes a sixth sense… you can sense the battle coming up, hoping you’ll make it to the save point or next town.
Dragon Warrior is a completely different beast. The phrase I used a lot at the time is that the frequency of random battles was “not just kindof random… REALLY random!” Of course, I know that this is not how randomness works, but it sums up the situation nicely. Sometimes, in Dragon Warrior, you want to grind but are forced to walk around for what feels like an entire minute before you enter a battle. Other times (typically when you are low on health and MP, as per Murphy’s Law), you get five battles in a row, literally entering the new battle after moving a single square after the old one. It’s infuriating, irritating, and absolutely exhilarating. I used to not use swear words… until I played DW and started screaming “f***********************k!” in my head when the battle-music started for the fourth time in four squares. Other things in the Dragon Quest series that are “really random” include the revive spell in later games (which, cruelly, only works 50% of the time, which of means you always have to cast it at least 5 times to get it to work), critical hits (the monsters get them when you have lots of health, but not enough to withstand a critical hit, and you get them when the monsters would have died by a normal hit anyway), and the actual in-game casinos.
Of course, the game is random in your favor just as often, but naturally the player never notices this. Except perhaps “that one time” when you have low health, no MP, and somehow manage to make it all the way back to the castle, and the one time you got into a battle with a Dracky you got a critical hit. Now, there’s obviously a bit of an issue with the use of random, turn-based battles in the first place, but if a game is going to have them I believe it would be best to imitate Dragon Quest.