Roguelike column thumbnail ['@ Play' is a kinda-sorta bi-weekly column by John Harris which discusses the history, present and future of the Roguelike dungeon exploring genre.]

I've beaten the drums pretty loudly for roguelikes here, I suppose. They are a style of role-playing game that has much to teach game designers, in its tactical depths, its subtle incorporation of logical puzzles, its open-endedness, and the sheer variety in them. As said last time, nothing in computer roleplaying games so preserves the spirit of 1974-vintage, Original Dungeons & Dragons as roguelike games do. I've cheered for roguelikes so much, in fact, that I believe I'm allowed a bit of pessimistic contemplation here. Roguelikes are great, yes, yes, but they are not perfect. There are some things lacking with them.

In this, I'm not talking about the kinds of things dredged up by many game reviewers when they're forced to comprehend a roguelike for the first time. Like the absences of: a central narrative; scripted events; clumsy metaphor; and amnesiac fourteen-year-olds somehow seeking absolution for their dark pasts while saving their generic fantasy world from empires fueled by the corrupting forces of Chaos with the aid of a dozen anime stereotypes.

Roguelike games do not lack for these things, because they never wanted them to begin with.

But if viewed in the light of classic Dungeons & Dragons, of a semi-adversarial game between players and a referee set in an ancient dungeon filled with all manner of tricks, traps and treasure, there remain some important things that roguelike games lack.

These lacks, nearly without exception, come from their following of Rogue. The ways that Rogue fails to measure up to the spirit of OD&D are also the ways that they fail. If one makes a study of the evolution of game design, one finds that this is completely common throughout the game industry. The first game of a genre gets its inspiration from some outside source; other people or companies, seeking to duplicate its success, make games that seek to improve upon the original by addressing its faults, but lacking the knowledge of the source material that the first game had; still other companies build upon those; and so on, until Strat-O-Matic sports has mutated into Madden '08.

But to return to the subject, these are all things that are not necessarily problems with roguelikes, but for some reason continue to be. These are things present in D&D that are not typically followed by roguelikes, and could conceivably gain by adopting. Well, I think so anyway:

D&D has much more interesting traps than roguelike games

rubegoldberg.gifD&D traps can be nearly anything imaginable; pressure plates, tripwires, magical detectors, monster cages, flame shooters, etc, but roguelike traps are uniform in comparison. The only way roguelikes can get away with this is because nearly every other CRPG has even less interesting traps, if they have them at all.

Rogue's trap model is very simple, and every roguelike I can name adopts it unquestioningly. Some of the spaces in the dungeon contain traps. The player is not shown those spaces at first glance, but if he steps into one it is revealed, and the player might be harmed by the trap.

Searching while on a space adjacent to the trap may reveal it, but searching every space takes too long and will likely cause the player to run out of food or die to wandering monster attacks, so in practice the player must forge on despite them.

In roguelike design theory, traps exist to provide a cost for careless exploration. It is impossible to entirely negate the danger from traps, so the player must instead reduce this risk by not stepping on unnecessary spaces, and retracing his footsteps whenever possible. This is a useful idea, but the game already includes the food requirement pushing the player towards move conservation.

In classic D&D on the other hand, traps are everywhere, and an integral part of the game. Often, especially in older modules, they can be deadly all by themselves. Traps are not placed willy-nilly but tend to guard things, like treasure or important rooms. And while surviving or defeating a trap is worth little or no experience points, any treasure guarded by the trap is worth experience in versions up to and including 1st Edition AD&D.

Exceptions: Nethack's rolling boulder traps are multi-part, and its bear traps and land mines can be disarmed, carried around, and even reset. These are recent additions however. The game still mostly follows Rogue in its trap model.

All monsters are, in tactical terms, the same size

In roguelike games, a pixie tends to be the same size, in terms of the game, as a dragon. Gnomes and giants take up the same space on the grid. They can all fit through the same doors, and down the same narrow passages.

Again, Nethack has an exception to this. It expended a good amount of code in making its long worm monster, which can extend for many spaces through the dungeon, and be split in two like an earthworm. Mystery Dungeon: Shiren the Wanderer, and some of the other Mystery Dungeon games, have some boss monsters that take up 2x2 spaces. Although those games aren't against conveniently forgetting this at times: their size doesn't seem to deny them access to narrow passages.

bigmonst.pngPossibly more roguelikes have not tried to simulate monsters more than one dungeon space in size because doors and passages tend to be one space wide, and it would limit their mobility. A monster bigger than that won't be able to leave the room it was generated in. In OD&D, it should be noted, passages are usually 10 feet wide, equivalent to two game "spaces," and nothing says the dungeon designer has to use standard doors for all rooms.

Hidden doors are all identical

OD&D had only three character classes: "Fighting Man," Magic User and Cleric. The final core archetype, Thieves, didn't enter the game until the first supplement. Before then, traps were detected and handled in what we might call "narrative space." The referee explains the surroundings to the player, and that description contains the clues that might lead the player to examine the trap's hiding place by investigating and asking about it, perhaps then disabling the trap by taking explicit action like pulling a lever or cutting a wire. (If this sounds a bit like playing a text adventure game, it's no accident. Colossal Cave was likely inspired by this style of Dungeons & Dragons, and Dungeon/Zork definitely was.)

When Thieves entered the game, they could find traps simply by succeeding on a die roll. This is the beginning of the game's trend towards including rules abstracting everything out. Finding secret doors, one of D&D's most thrilling discoveries, also comes down to a die roll. Instead of pressing a loose brick or finding a hidden lever, the player simply must search. Roguelike searching works the same way.

(A thread on explains a bit better what this kind of trap description adds to the game.)

Stairs not matching X and Y positions between levels

A fact I find interesting: no roguelike game I've seen even attempts to have stairs and other ways of going between levels be logically consistent with each other. Unlike D&D dungeons, stairs on one level almost never match up with the same location on the floor below. In other words: if you find two stairs next to each other in a game with multiple stairs (like Angband or Dungeon Crawl), if you were to go down each of them, you would probably end up in spots far removed from each other on the next level. Stairs don't preserve X and Y location when taken, they instead come out anywhere on the next level. And pit traps lead to a random location on the next floor that changes every time they're fallen down.

This may seem minor, but it prevents the level generator designer from including intra-level mazes, like places where the player must travel back and forth between two levels to ultimately find the way down to still deeper areas, or secret rooms that are intended to be found by finding stairs from the prior level that can be found by paying attention to the level.

There is a technical reasons that roguelike games do this. Rogue was created at a time when computers had very little memory, so levels are generated when entered instead of at the start of the game. Later games aren't so restricted in memory, and some games like ADOM, Crawl and the 'Hacks either save levels not being explored to disk or keep them in an alternate memory buffer, but they still generate the levels on a need-to-see basis.

While the initial entrance to the level could be guaranteed to be a suitable location for stairs (that is, not in a wall), if the levels were to get generated in a non-linear manner, inconsistencies could arise. If levels 15 and 17 get generated before level 16, then whichever entrance to 16 is used first could be guaranteed to get a consistent staircase, but the stairs to the other level might not. Forcing staircase locations before the map is created means that level must be build around them, and random map generation is still enough of a black art that this might be regarded as being too difficult.

Now of course, this is only the case if the whole dungeon isn't created at the start, when these matters could be resolved simply. And even if the game doesn't keep the whole dungeon in memory at once, a sufficiently-advanced generation algorithm certainly could ensure logically consistent stairs. That no roguelike to my knowledge, and certainly no major game in the genre, has done so yet is strange.

Item identification is much easier than classic D&D

One of the most interesting things about roguelikes is how item descriptions and types match up. All items of a given type share a description. Bubbly, yellow, milky and purple potions are different types, types that might not be known at the start of the game but persist during it. All types of one description are the same, so once one is known, they're all known.

For Rogue the item system is well-balanced, for there practical limits on the number of items that could be generated on each level, there were no guarantees that any specific item would be found, and even if the player managed to explore down to Amulet level, chances were high that some items would never make an appearance. The result is that, even at the deepest levels, players continue to find unknown objects, good or bad, whose types must be discovered through expending resources, insight, process of elimination, or risky usage. And since no item is really essential to win and there are tricky uses for some bad things, even an exceptionally unlucky player who only finds bad stuff still has a chance.

This balance has changed as roguelike games became more complex, and dungeons increased in size. In a game of Nethack, nearly all random items will be generated, either randomly or from guaranteed creation, before the player reaches Gehennom, halfway through the dungeon. Those items which are certain to show up do so because a few kinds of items are necessary to win: a means of crossing water, for example, will probably be needed on either the Medusa, Castle, Juiblex or Plane of Water levels. On the other hand, successful Angband players spend lots of time farming levels in order to find the stuff they need. Angband changes item generation probabilities as the player gets deeper into the dungeon, so not only must he know what he needs, he must know the levels it's found on.

The roguelike identification system was inspired by Original Dungeons & Dragons, in which items were much harder to figure out. Many items had secret switches or command words that activated their powers, and did nothing until those were found. Cursed items had much more variety, and were much more diabolical. There were a number of extremely bad items which looked exactly like more valuable stuff, like potions of delusion. A bowl commanding water elementals is a very useful object, but it looks exactly the same as a bowl of watery death, which can kill a character so dead that even a wish won't bring him back. These items seem to exist only to punish players who might try to figure things out through observation! Enough referees must have seen this as unfair that most recent versions of D&D de-emphasize the identification of treasure. In an OD&D game he ran in 2005, Gary Gygax himself told his players what most of the stuff they found was, although there were some exceptions to this.

Whether this aspect of Dungeons & Dragons, that even with experience there are items that are fatally indistinguishable from each other, is good or not I don't think I can say. I recognize why it's there, but I lean against it. Item identification has become a uniquely distinguishing characteristic of roguelike gaming. And yet, the identification game of roguelikes other than Rogue tends to get easier later on to the point of meaninglessness. Especially in Nethack: most of its items will be discovered before halfway through the dungeon, and this is unquestionably one of the sources of that game's declining difficulty midway through. Fixing this problem would require design innovation, but there's no reason to believe it's impossible.

Lack of multiplayer

multiplayer.png Adding multiplayer to (insert roguelike name here) is one of those features that seems so obvious, but makes their programmers weep openly. The Nethack Dev Team has a whole item on their FAQ devoted to rejecting this idea. The problem is that roguelikes are turn-based. If you're in the same room as another player, and he nips off to the bathroom for a while, what do you do in the meantime? Twiddle your thumbs?

The reason multiplayer would be important to roguelike games is that it is a feature of pen-and-paper roleplaying that is lacking in CRPGs: the participation of other players in the game making it into a social experience.

It is the one aspect of old-school roleplaying gaming that another genre arguably does better than roguelikes: MMORPGs may have plexiglass-coated designs that reduce player classes to a surprisingly limited set of archetypes, prevent characters from ever dying or even suffering any serious losses, and typically have designs that center around combat as the be-and-end-all of RPG play, but by gum, they do let you cooperate with your friends.

To their credit, MMORPGs actually solve it fairly well. The biggest change they seem to suggest is that a multiplayer roguelike should be mostly real-time. But in addition to the fact that this would completely change the way a roguelike game is implemented, it'd mean that players would no longer be able to sit and fully think through the ramifications of their moves before making them. These unexpected ramifications are a major part of the appeal of many roguelikes.

So fixing this problem is not a trivial matter. But it could well be solvable. In fact, it may have already been solved: the Angband variant MAngband is a full, internet enabled, real-time variation of Angband. I haven't looked into it much yet, but it's been around for years and is still undergoing active development. It seems to have an active player community too.

Many years ago there was a project called Interhack, which purported to be a multiplayer recreation of Nethack. Its developer claimed to have devised a system he called "surreal time," which shifted between completely turn-based and real-time play depending on how close players were to each other. Alas, it seems its developer became disenchanted with the project along the way, or became involved in an altercation with some of the Nethack player community, and removed all traces of the project from the internet. But undoubtedly, there are important lessons there for roguelike developers contemplating allowing more than a single player into their dungeon at once.

Wired's Game|Life recently reviewed the new Pokemon Mystery Dungeon games. While I agree with the idea that the game is tired in that the games are dull (as I remarked before), it should be noticed that this is a direct consequence of the supposedly wired item that it's a roguelike that's not punishing. If only I could shout this through some kind of Internet loudspeaker so I wouldn't have to explain it yet again: a roguelike game that isn't dangerous is pointless. Without a strong narrative to carry them, they amount to just bumbling around and hitting stuff until they die – like nearly every other CRPG out there, but without even the pretense of advancing a story.

The sooner this basic, but essential, concept is learned by both players and reviewers, the sooner developers will be able to abandon attaching yawn-inducing narratives to them, and be able to devote their energies to making their games deeper. Which, if they're going to compete with open source computer roguelikes, they are absolutely going to have to do.

This pretty much concludes our examination of roguelikes in the context of Dungeons & Dragons. Next time out on @Play, we start examining the games of the Roguelike Restoration Project, by taking a look at SuperRogue. See you in two weeks, more or less.