COLUMN: @Play: The Eight Rules of Roguelike Design

['@ Play' is a monthly column by John Harris which discusses the history, present and future of the Roguelike dungeon exploring genre.]

Back in November, in the previous @Play column, I mentioned a number of proposed rules of roguelike design, and promised soon to describe them. It's taken a bit longer than I expected, but here they are.

I call these rules for rhetorical purposes only. I don't think there are any inviolate laws of game design. But given we are talking about roguelikes, there are certain properties that have been important to the genre.

Maybe not to all roguelike games; some of these have to do with designing a good item identification system, for instance, and many of the more recent games do not use that. I'm fairly outspoken in my appreciation for item-ID systems, so please calibrate your wonk-o-meter appropriately.

I use the term "reasonable play" several times here. It refers to being in a neutral state in terms of danger. For example, almost any bad effect from using an item can prove fatal if the player uses it at the wrong time. Discovering the potion of confusion by quaffing when a troll is attacking you is dangerous—so don't do that!

f you're down to one hit point, even slight damage could kill you, and some games have items that do piddling damage, so don't do that either. Most of the time in a roguelike the player is not in immediate danger. That is the fact on which the idea of reasonable play rests. Unknown items are possibly dangerous, so there must exist times of lesser danger in which to try them. A game built on the idea of literally constant peril would have different design demands. You are on your own in figuring those out.

Now, while a bad effect is active, it's possible for a previously out-of-sight monster to walk up and start hitting the player. That is a less obvious case, but the player could have test ID'd the item while in a large, lit room, decreasing the chance that an unseen monster could reach him before the potion wore off. This helps to make it okay. In fact, the most effective roguelikes purposely make it difficult to always know when it's safe to perform dangerous actions. If there were a hard-and-fast rule about test-IDing potions of confusion, then an argument could be made that they shouldn't be in the game!

Of course game design is not a science. I tend to rebel whenever I hear someone tell me about features that "obviously" should never, or always, be present in a game. The following rules are no different. They are useful, however, when talking with respect to the default state of roguelike play and design, which I will define here as being that of Rogue itself. But they all apply, with different strictness, to all the other major roguelikes: Nethack, Angband, ADOM and Dungeon Crawl.

So here is a list of eight rules. Each leads with a name, in quotation marks, to facilitate discussion, followed by the rule itself in around a sentence, followed by discussion, and then finally followed by both examples of games using it well and "reverse examples" of games doing it badly. I apologize in advance for there being a lot of Nethack in these examples, but, well, it still has many features worthy of discussion.

1. "No beheading rule." Provided reasonable play, the player’s character should not be killed or harmed too greatly & permanently in one attack.
Example: In Nethack, cockatrices can't immediately kill the player through a single attack. They can initiate delayed stoning, but that gives the player a few turns to cure the condition.

Reverse example: In old versions of Nethack, Medusa was a random monster that appeared in a random room in the deeper dungeons. Since merely seeing Medusa kills the player, this breaks the rule unless the player had a way of knowing Medusa was there before stepping into sight, and the game was random enough that there was no reasonable chance of that happening. This was arguably bad design, which may explain why more recent versions of Nethack put Medusa on a special level, where at least experienced players will know where she lurks. Still, Medusa always appeared on a specific level of the dungeon, and it was always on the downstairs which were never generated in the same room as the upstairs, so a player could conceivably be prepared for her. It's still rather more reliance on spoilers than anything in Rogue requires.
There also items that can be wielded by monsters that can kill with one hit, such as Vorpal Blade or the Tsurugi of Muramasa. But the chances of running into a monster wielding the first in the dungeon are vanishingly rare, and the second is in a specific location like Medusa, and only appears in the Samurai quest anyway. There are also other difficulties with monsters with wands of death, Touch of Death spells and disintegration beams and the like; after a certain point, the game requires that the player be proactive in protect himself from sudden instadeaths.

Reverse example: Shiren the Wanderer. Shiren is mostly really good at this, but its Skull Wraiths may be a little too strong. Their distance attacks can inflict amazingly dangerous status ailments that could conceivably result in Shiren dying before he gets to make another decision. With careful play this can be negated mostly, but there is still a chance on the last wide-open level that the player could face a situation he cannot escape.

2. "No cyanide rule." No unidentified item should be immediately fatal upon use given reasonable circumstances, otherwise the player should never try to use-identify.

Example: In Rogue, probably the worst item is the potion of blindness, which makes the game nearly unplayable. Not only does it give all spaces on the level a vision range even worse than that of dark passages, it is impossible to find secret doors while blind, which state has a good chance of stranding the player until the blindness wears off of he drinks a potion of extra healing. But there is not ring of blindness, because for such an item to have bite it must be initially cursed, and if the player put on a cursed ring of blindness he might not have any means of curse lifting available, and if that were true and he was blocked from the exit by a secret passage he would not be doomed to starve. The effect would be to make it unwise to try to test-identify rings because of the chance of death. One might qualify this by saying it'd be unwise to test-ID rings if no means of curse removal were available, but considering all curse-lifting in that game is from random sources it might be considered unreasonable.
Example: In Nethack, there is a random item that can kill in normal situations just from ordinary use: the Amulet of Strangulation. However, it is a delayed death, and prayer can get the player out of it. However, what if the player has recently prayed and his timeout hasn't expired? What if he's in Genhennom and can't effectively pray? What if his luck is negative, making it impossible to pray effectively? These aren't the more common states though. In Genhennom, the player should know enough lore to know not to test amulets there.

Example: There are other items in Nethack that can kill instantly from use, but not by their most common uses. The Wand of Death, if zapped at self, kills instantly if the player isn't magic resistant. But players don't usually zap random wands at themselves, so one might consider this to be reasonable. Reverse example: Note that there _are_ cases in Nethack where death might be considered unreasonable; every random egg has a chance of being a cockatrice egg, so eating eggs is an unexpectedly bad idea.
Reverse example: Dungeon Crawl has a potion of poison with a long-lived effect. This one plays carefully with the line, since it could well prove fatal to a very low-level character, but usually by resting during that time it has a chance to wear off before death ensues.

Rules 3 and 4 are about item identification. Many recent games don't include this feature. Here I give my reasons for suggesting such a system. While item identification in roguelikes does aid in presenting the world as a mysterious place to figure out, and in giving the dungeon itself a dangerous character as opposed to just the monsters, there is also a game design reason. It disassociates the player's progress from the random number generator: the player could find all of the best items in the game on the first level, but he'll still have challenge figuring that out without wasting the resources. The items found are not a gift to the player; he must still expend some effort to properly understand what he's found.

The identification game in roguelikes is one of the least considered by designers. Even to this day, it's probable that Rogue has the best ID game, although Nethack's isn't bad (although mostly irrelevant as soon as the mid game). Rogue's is designed so that scrolls of identify cannot be relied upon, for players almost never find all the scrolls they want. Plus, since no random items are guaranteed, the player never knows if he'll find another item of a type. A few items are extremely good but easy to waste. The best of them all (scroll of scare monster) shows unusual care in its design; using it in fact wastes it, but it is the only scroll that can be identified without even picking it up.

3. "Item masquerade rule." Items should be difficult to identify even for spoiled players, due to similarities between items of the same type.
It is best if this rule is not perfectly adhered to; giving observant players some benefit for their insight is in keeping with roguelike gameplay.

Example: Yes, another Nethack example. Both Potions of See Invisible and Potions of Fruit Juice generally have the same messages upon use.

Example: In most games, the effects of most rings are vague and can only be deduced through close observation, if even then.

Example: Nethack and Shiren both take steps to obscure item identification through shop pricing by offering many items that sell for the same amount. Example: ADOM clearly marks the weights of items on the inventory screen, in "stones." A few random items and artifacts have distinctive weights, which an observant (or heavily spoiled) player can note and use to pick those things out.

4. "Situational ID advantage rule." When unknown, item effects in a category should overlap in such a way that use in some situations would be good to discover, while others simultaneously would be bad.
If no potions provide combat disadvantages, then it's in the player's best interest to test them out while fighting. If none do direct damage but one does heal, then the player should only test ID when below maximum hit points. It is easy to go overboard in eliminating these kinds of situational ID advantages, but a good designer will still be cognizant of them and devote a little thought to lessening their influence on the game.

Example: Rogue has items that can be very useful when used in battle (extra healing) and very bad (confusion, blindness). When zapping wands at monsters, it's possible a monster to be slowed (very good), hasted (bad), teleported away (generally good) teleported directly to the player's side (somewhat bad) or polymorphed (potentially game-ending at early levels, more beneficial as the game continues). Study of the items included in the game make it clear that some thought was put into providing a healthy amount of risk in most situations, to provide fewer "no brainer" situations to using unknown items.

5. "Item enchantment rule." When known, items should as much as possible present interesting decisions.

Example: All of the major roguelikes, to some extent, use scrolls of enchant weapon and enchant armor to do this to a degree. When you use a scroll, its enchantment bonus becomes a permanent fixture on the one item it's used on. If a better item comes along later that the player would rather use, maybe one that already has a high enchantment, it could be that the scroll was wasted. However, the longer the player waits to see if something better comes along, the less total use he gets out of it. All the major roguelikes feature this to a degree, but it's Rogue that makes the best use, since as stated previously, the player has no idea if he'll find a better weapon or any more enchantment scrolls.

Example: Also in Rogue, armor has an extra wrinkle. The weakest, Leather Armor, has the special property of being immune to the rust caused by rust traps and rust monsters/aquators. If the player uses all his enchant armor scrolls on leather, he can be sure of never losing those enchantments, but he probably won't be able to get the suit enchanted up to the heights of plate mail. The player can overcome this a bit waiting until after the aquator floors before using his scrolls. All versions of Rogue offer rings of maintain armor which allow the player to not worry about having his armor weakened, though at a cost of food.
Reverse example: Later versions of Rogue also provide scrolls of maintain armor, which, if one appears (they are fairly common) make the decision pointless.

6. "Two-sided coin rule." Given perfect knowledge of identifications and uses, items should never be completely useless.
In fact no roguelike that I know of does this perfectly, but many of them show a lot of effort has been expended towards reducing the prevalence of completely useless items. The reason for this is not just to reward player wit and knowledge; if no item is entirely bad, then the player is additionally insulated from the caprices of the random item generation. This does not mean that all items need be great.

Example: In Rogue, even potions of blindness are useful when fighting medusas, which sometimes confuse on sight. Bad potions can also be thrown at monsters to possibly affect them.

Example: In Shiren the Wanderer, most bad items will affect enemies when thrown at them, making few things completely baneful. This can even be used to get a last use out of depleted wands.

Example: Nethack is pretty good about this, and has shown special care in making sure many bad states have hidden upsides. Hallucination protects against "touch of death" magic. Confusion is sometimes very useful, providing most scrolls with alternate effects (which also makes some bad scrolls into good ones). Thrown potions break and provide vapor effects (which are, sadly, usually negligible). "Bad" wands that help the monsters can be zapped at yourself to give you their benefits. Conversely, there are also good items, such as Wands of Cancellation, that can be very bad if used incorrectly.

7. "Reducing grind rule." The mere act of spending time should deplete some finite resource, forcing players to keep those resources up, while also limiting the amount of "grinding" he can do.

Example: Historically, this is done through a food timer. Rogue has the best of these: food only ever enters the game through item generation, which is tied to generating new levels. Technically food can be created by killing some monsters, but that's fairly rare, random, and suitably dangerous.

Example: Dungeon Crawl's food timer, while not as strict as Rogue's, still serves its traditional function of forcing a hard limit to the player's activities. Dungeon Crawl is explicitly designed to reduce grinding, an admirable goal.

Example: ADOM does this okay for the most part. Although food is not limited as strictly as in Rogue or Crawl, its corruption timer serves much the same purpose. There are loopholes in its item scarcity, however, that can give particularly clever players a way out from under the weight of the clock. Look up "ADOM wish engine" on Google sometime….

8. "Race you can't win rule." The game's monster difficulty should increase slightly faster than the advancement of the player, given average stats and default equipment, so as to force him to rely upon items and tactics.
The reasoning here is that if the player doesn't have to rely on randomly-found stuff then they become unimportant to play. However, if it's required to have specific items to be successful then many games will be outright unwinnable. The balance between these two poles is what makes random dungeon generation difficult, but it's also part of what makes random dungeon gameplay interesting.

Example: Rogue is, again, best at this. Even with excellent equipment, you can't just pound through all the monsters; starting around the time you encounter trolls, you'll need to use some tactics or resources to survive more than one fight, and griffins up the ante even further, not to mention medusas, ur-viles, and of course dragons. Shiren is also pretty good at it; more enemies in that game have special attributes, which in practice means you can get further on strategy.