COLUMN: @Play: Purposes for Randomization in Game Design

['@ Play' is a monthly column by John Harris which discusses the history, present and future of the Roguelike dungeon exploring genre. This time -- a look at how randomization of game elements - so popular in Roguelikes - works to draw in the player and provide replayability.]

Roguelikes are the genre of game most associated with randomization of game content. Some other games, especially classic games, feature substantial random elements, but most of them use them as an afterthought, or in an ineffective manner, or without fully realizing how they change the dynamics of their game.

As an example I offer a Gamecube game, Nintendo's Pikmin 2. One of the advertised features for the game of the randomized dungeons that players would enter during play. I consider that the original Pikmin, which had little or no random elements, to be brilliantly designed.

It forced players to make wise use of their actions within a strict time limit (30 game days), and yet the limit was short enough that mistakes were duly punished and players had a disincentive to grind for additional troops. Yet, it wasn't really so difficult that it took players more than one or two attempts to win. And really dedicated players could win surprisingly quickly; the shortest possible time to win was nine game days out of the thirty allowed.

But some players, and more importantly game blogs, complained about the overall time limit, so it was removed in the sequel. Further, for the random dungeons which were advertised as the major addition to the game, the day timer is disabled. Each floor's layout is not actually randomly generated: instead, the maze layout, the treasures to be found, the resources provided, and the enemy opposition are all static, predetermined for each floor. Without the timer, there is no pressure that might force players to move on without having found everything.

The only real randomness is in the locations of all these things. Scrambling the start location too helps a little, but not much. Thus, the only effect that the randomness has on the game is the order in which things are found and the relatively small variable tactical challenge that comes from fighting enemies under differing terrain conditions. For all the promise that randomization maps held, it really barely made any more sense to scramble object locations than to keep them static too.

When does it make sense to randomize? What benefits does this provide to a computer game? Here are a few. (I should warn you ahead of time, this is one of the more "out there" columns in the sequence....)

Preserves player interest across multiple playthroughs

This is the reason most given, and it is not a bad one.

Most games will always have the same monsters in that corridor, or the same item in that treasure box. Playing through it a second time provides little in the way of a new experience, since everything is as it was the first time through. It's not that a game cannot be played twice, it's just that, the hidden knowledge of the game having all been revealed, a fundamental element of the experience is missing. The decisions the player makes in progressing through do not change.

He may optimize his playing style in order to best make use of the hidden events coming up, but this is a process substantively different from really playing the game. It is the difference between playing Super Mario Bros. to get through it and trying to speed run it; the game ceases to be about reacting to unknown events, an extremely complicated task requiring a human being to overcome, and becomes about planning and executing a trip through the game, which is difficult for entirely other reasons. It is the difficulty of forming a plan, and of exactingly following it, but not of reacting to unexpected events. Randomizing a game can provide this.

But allow me to pose you a question. What if you had a generic Dragon Quest-style JRPG in which every aspect was the same every play through except the layouts of the dungeons? Even with the dungeon contents the same, just mixed up every game? Would that game, in truth, be any more replayable? A player may wander down dead-end hallways and thus fight a similar number of monsters to his first time through, but would a player really find that interesting enough to bother playing through a second time? A hypothetical roguelike of that type, which had everything static between plays except for dungeon layout, would be a little better off because the dungeon creates a tactical combat challenge which may vary between game. Even so, such a game would not be all that entertaining.

Provides resistance to spoilers

This is similar to the first point, but increasingly important with so many FAQ sites out there and increasingly many players who are happy to steamroll through a game.

It is obvious that a game in which large portions are randomly generated is at least somewhat resistant to FAQs, but it is a difficult task to make a game that's spoiler-proof. It is impossible to write a walkthrough for Rogue, but on given floors certain monsters tend to show up, and there are learnable, and thus spoilable, strategies for how to deal with them. There are only a limited number of possible items that can be generated, and both learning what they are and using them are vulnerable to FAQs. Nethack takes this to extremes; it's almost impossible to win without either spending years learning the game or reading a truckload of FAQs, but once you are sufficently spoiled winning the game is much easier than you'd think. You might not be able to win every game like Nethack god marvin, but winning one in twelve is definitely within the reach of the average gamer, and more importantly once you get past a certain point early on winning can be almost assured. (In my opinion this is the point where you've gotten poision resistance, magic resistance and dragon scale mail.)

I doubt it possible to take game randomization to a point where FAQs and other gameplay spoilers become useless, at least before the point where computers will be able to generate entire games whole cloth, which is several decades down the road at least. In the meantime, the only real resistance to spoilers there can be is to make the act of playing the game more fun than getting to the end. That would remove the incentive to cheat through the game.

A component of the urge to get through the game as quickly as possible and by any means necessary is the tendency of many developers to put in “unlockables,” content that the player must earn the right to play. (That is ultimately an unhealthy development; for proof of this statement, I point to those lamentable games where the “unlocks” are actually DLC, a situation which would have made less intuitive sense to publishers if they hadn't been locking off features from players since the NES days.)

Challenges player deductive skills

The combination of randomly generated and predetermined content, if planned right, can make possible an interesting kind of gameplay, one of the few that allows players to make use of deductive skills. Roguelikes generate random items from a list of possibilities, and identifying them requires making use of observable evidence in order to figure out item types while reducing or eliminating the risks of use.

Nethack makes the most use of this. The Dev Team went so far as to group items according to sale price, with most price tiers containing at least one bad item, to reduce the utility of “Price ID” as a means of item identification. (Observant players will note that the Shiren the Wanderer games group item prices in a similar manner, another proof that Shiren's Japanese developers must surely be aware of Nethack.) Overall, however, item identification is somewhat out of fashion as a means of adding challenge to a roguelike game; Dungeon Crawl's developers have outright stated that a design goal of theirs is to reduce the importance of the identification subgame, possibly because to make use of it the player must have a list of possible item types at hand, which to utilize demands either extensive note-taking over many games or spoilers.

It is possible, although uncommon, for a roguelike game to use this design concept in areas other than item identification. One of the few examples that comes to mind is Nethack's endgame, in which players must discover the proper altar, out of three possibilities, on which to complete his quest. Also in that game, it is possible to figure out what the reward is for finishing Sokoban before completing the puzzles leading up to it, which can aid an experienced player in deciding whether to go through with it.

Escapes the limits of conventional narration

This one has only recently become solidly recognized. Yet it should not be underestimated; Dwarf Fortress has utilized it to become a darling of the internet. I should warn you, I'm going to display my under-utilized English Major powers here. If you are allergic to high-falutin' discussions on the nature of art, you may want to move to the next section heading.

The practice and technique of storytelling has evolved over thousands of years. Since Aristotle, all kinds of people have weighed in on what a story is, what it should be, and what makes them good. When game developers decided they wanted it to become a storytelling medium, they adopted that advice and formulated their own rules, some unconsciously, as to what such a game should be. Rules like a three-act structure, rising action, a climatic moment, a paring away of irrelevant elements, a unity of theme, and a certain purposefulness, sometimes going so far as to provide a distinct moral. These rules, in many ways, stand opposed to the ideals of game design.

Storytelling was devised to describe life, imposing a narrative pattern upon a series of real-world events. Fiction takes that pattern and seeks to synthesize a series of events to match it. Games also seek to emulate life, but whereas a story attempts to make sense of the randomness of life by suggesting a plan, games revel in that randomness.

You don't know how a (traditional) game will turn out, and that's the point. When you start a game of Monopoly, when you take to the field in a game of baseball, when you begin a hand of Bridge, the idea is to overcome a challenge set before you, one with inseparable random elements, through skill and wit.

That challenge may or may not conform easily to a narrative sequence. Sometimes a player heroically overcomes great odds in order to prevail, but usually the great odds win out. The odds-on favorite usually wins. The many times when the expected result occurs would be considered wasted to a storyteller, who instead of describing everyone the Cyclops ever devoured before Odysseus showed up wants to get right to the good stuff, but to a gamer they lend weight to those times when the expected results fail to occur. And when the player can increase the chances of success through his actions, then he can feel good about himself because his skill really has won the day.

In games with pre-written stories, the favored result rarely happens. The evil forces with so much going for them rarely win out. Even when they do, it's not the fault of the player one way or another. Even if the player loses (or “gets a game over” in the parlance), he can return to a previous save and try again. Players subconsciously perceive this fact and it robs video game storytelling of weight.

Whether the ending to a God of War game has Kratos prevailing or not, it's already been determined which it is; the fact that the player must walk Kratos through the steps to get to that point doesn't affect his chances of success, they just slow down the story. Because of this, I say the narrative and the gameplay are not well-integrated. They are created following techniques for writing fiction: starting from the story and inventing the details to fill the mold. And anyone who has ever watched Scooby-Doo knows that this has certain drawbacks if the writer wants to keep the players in suspense.

Roguelike games have, for a long time, been one of the few types of computer games to go about writing the other way around. That is to say, few of them seek to write anything at all! They seek to emulate a realistic world, with enough tricks and traps thrown in to make it interesting, and ask the player to write a legend of his own. The legend means something because it was purchased with the player's skill and sweat; the story of the character becomes the player's own. Dwarf Fortress (the primary mode of which is not a roguelike, but it is kind of an honorary example of the breed all the same) makes this almost explicit, to the degree that there is a whole category of DF enthusiast: one who will never brave the UI or learn how to construct a moat, but loves to read the play reports, a.k.a. legends, of fortresses like Boatmurdered, Syrupleaf, Headshoots and Bronzemurders.

Why are video games worthy to be considered an important art form? What can we use to counter Roger Ebert when he claims so certainly that they are not art? My answer to him has nothing to do with providing an experience to the player. No, ultimately the future of these games lies in their use as a creative tool, a way of exploring possibility, a method of aiding the player in creating his own narrative. In this, computer games have the potential to touch the divine in a way no other work can. It is meta art; art that aids in the production of other art. Every other art form limits the audience's involvement to interpretation; only interactivity can draw him in beyond that. In this area video games are unique, but the great majority of games squander this potential shamefully. But the best roguelikes do not.

Provides complex play through combining of basic elements

Let's end with a more solid use for randomized gameplay. It is also the one that is the most solidly roguelike: using it as a means of throwing together different features to provide unique adventure experiences.

For an example, let's take Rogue. There are no character classes in Rogue. You always begin with 12 hit points, 16 Strength, a +1, +1 mace, ring mail, a shortbow and arrows and one food ration. At the moment of starting an game, your adventure always begins the same way.

Rogue is balanced, generally, against the player. At the start of the game the player is only a little stronger than the toughest type of monster he may face on level one, Hobgoblins. He gains levels rapidly, but a new type of monster is introduced every floor until level 23. Given average rates of decent and combat, monster difficulty keeps pace with the player at first, and eventually surpasses him. And the hunger limitation forces him to keep exploring new dungeon levels instead of grinding out experience on the easier floors.

The only thing that gives him a chance of surviving, maybe even prevailing, is the random treasure generated along the way, and because the player is doomed without it he must make the best use of it he can. Many of these items give the player a minor advantage. Stronger armor reduces damage taken in combat; stronger weapons mean fewer successful strikes are needed to overcome a foe; wands provide a limited resource for overcoming the monstrous opposition; rings of maintain armor protect your armor from damage, but consume food while wearing it; leather armor is weak but cannot rust and doesn't take up a ring slot; on the other hand rings of slow digestion push back the spiked wall of the food system a bit, letting the player level up a little, but in exchange for taking up a ring finger that might be used for something else.

Many of these items change the game significantly, even radically. Careful use of a wand of slow monster, for instance, can defeat most monsters, even if they greatly outpower the player. But a single magic item isn't nearly enough to get the player through the deeper levels; it is by using several at once, and taking advantage of the ways that their powers overlap, that will see the player through.

If it wasn't for the various ways the treasure changes the game, Rogue would not be nearly as interesting to play. Merely having random dungeons and monsters is not enough to make it interesting; dungeons ultimately are just differently-shaped spaces to explore and empty space isn't that involving. Monsters are a more interesting, especially with all the special powers Rogue's set enjoys, but generally the same kinds of monsters are generated on each floor, so you end up facing everyone eventually, just in a slightly different order and proportion. If you can't handle trolls this game, you probably won't be able to handle them the next either.

But, what if you attacked a troll with a ring of increase damage +2? Or wearing plate mail? Or with a ring of regeneration, to match that of the troll? Or what if you have a scroll of scare monster? Or what if you use a potion of monster detection to see it from halfway across the map, and thus know where to avoid?

Collectable card games use a similar premise in the way played cards can modify the rules of the game. If the same cards, or items, could be drawn/generated when desired every game, then they would cancel out of the rules. Many games which allow for player choice at the beginning of a game to choose which options to take, when played by a sufficiently knowledgeable player, boil down to exactly this if he has determined that even one such choice is more powerful than the others. Then it becomes like those other choices aren't even available; why would a player choose a weak benefit if an obviously better one exists? Obvious choices are not really choices at all. In these cases, good design would demand that the designer just make them part of the main rules and not ask the player to much a trivial decision.

Making the player's advantages random forces him to make use of sub-optimal equipment at times. Rogue does this excellently; you probably won't find every item on a single playthrough, so you make due with what you get. Nethack is arguably worse designed than Rogue in this regard, since practically everything will be found on a given run-through, and if the most important stuff isn't by some chance, there are guaranteed items around to make up the slack, and a wand of wishing in the event of a hostile random number generator. These safeguards remove randomness from the game; you need never worry about a game in which no boots of levitation are ever generated, because there is a guaranteed pair in the game. Nethack does save most of its essential equipment for late in the game, but the general principle remains.

These failsafes make Nethack fairer, but at the cost of making it less interesting. That these two qualities exist, to some extent, in a mutually exclusive relationship, where reducing one increases the other, is strange, yet perhaps revelatory.