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

This is the beginning of a sequence of articles on the popular roguelike game Dungeon Crawl. We've covered it once before, but considering the game's importance and continued development we have not discussed it nearly as much as it deserves. Hopefully this and the next few articles will go some way towards remedying this tragic situation!

Of the five major roguelikes (Rogue, Nethack, Angband, ADOM and Dungeon Crawl), Crawl is both the most recent addition the list and the one undergoing, by far, the most intensive development. A favorite of the Goons over at Something Awful, it possesses a very strong design which is difficult to exploit, and provides tradeoffs and drawbacks for most important actions. In this it sticks closely to Rogue, and other than the original Hack it is probably the popular roguelike that best recognizes its forefather's great strengths.

These articles are written based on the as-of-this-writing most current stable version of Dungeon Crawl Stone Soup, 0.5.2. Much of the information herein was gleaned through perusal of the Dungeon Crawl Wiki at, and the spoilers found at It should be noted that a new version is under development as v0.6.0, and that a development build of this version is available for download.

Crawl’s experience system

To begin our discussion, I must first describe Dungeon Crawl’s unique experience system.

Characters in Crawl advance in two important ways. The first way is the usual one featured in most role-playing games, through the gaining of experience points towards raising character level from defeating monsters. Crawl is more conservative in its advancement rules than some other RPGs; unlike Moria and Angband or later editions of D&D, monsters do not grant less experience when they are outclassed by the player. The reduced benefit of defeating weak monsters comes entirely through the greatly-increasing experience needed to gain higher levels.

But while “character level” in Crawl has certain visible benefits (HP and MP go up, stat increases may occur, and more spells may be learned), a character tends to gain more profound advantages from increasing skill levels.

dcss1.pngCharacters have a number of skills, which can be examined in-game by pressing the ‘m’ key. These can be divided into fighting skills (which includes general fighting, all the weapon skills, and some additional abilities like Dodge and Throwing), utility skills (Traps & Doors being the most universal, but also the specialized magic-using abilities Invocations and Evocations), and magic skills (one per school of magic, and general spellcasting). These skills are what govern most of a character’s action success rates and power.

When the player earns experience in Crawl, both the overall experience total goes up and an extra total called the experience pool. Points live in the pool until they get assigned to one of the skills through practice. Pretty much any action the player can perform that has a chance of failure belongs to one of the skill categories, and practicing that action provides a chance that some of those skill points will be moved to that skill. Higher skill levels require vastly more points than lower levels, just as with character level, but will also increase the number of points that are diverted to that skill with a single practice.

The most interesting thing about Crawl’s skill system is how it deftly avoids the many problems of a skill-based development system.

  • - There is no feeling a character doesn’t improve over time,
  • - nor that skills are watered-down to prevent one from being unbalanced compared to the others
  • - nor that one is overpowered
  • - nor that characters are generic, infinitely-malleable bags of properties interchangeable with each other
  • - nor that a character quickly maxxes out and becomes unable to advance
  • - nor that rapid, early advancement in one area makes improving in others impossible or nearly so,
  • - nor that a character can easily advance in all areas and become a super character without proper effort and playing skill.

Crawl’s system avoids all these problems. It is probably the best skill system yet seen in any roguelike; it could make a claim at being one of the best in any CRPG. Aimless practicing only improves skills so long as the player has successes, measured in monster kills, to power them, and also requires an opportunity to train them. The skills are broad enough that no experience diversions feel wasted, but sufficiently narrow that characters don’t become, functionally, generalists. If a character has low aptitude in a skill it costs more points to train, a significant but far from insurmountable amount. And most skill practice events require some opportunity, meaning even if the player has pool points to spare, there may not be the opportunity around that the player wants to spend them on.

(Oh by the way, the maximum experience level, skill levels, and dungeon level of Dungeon Crawl all are 27. Don’t ask why; it’s kind of the magic number of Dungeon Crawl.)

Race, Class, Stats and Deity: the four facts of Crawl existence

A starting Dungeon Crawl character is, generally speaking, a combination of race, class, statistics and god.

Races are a bundle of skill aptitudes, some starting equipment, skills and stats, and sometimes a few special rules. Aptitudes determine how difficult it is to train a specific skill. Special rules are sometimes profound changes to the basic gameplay of Crawl, like an inability to use most armor or a rapid metabolism. A hill orc character, of whatever class, will have certain abilities that a deep dwarf of the same class is not capable of and vice-versa. Many of the game’s races have special abilities like this. Some of these differences may overcome with the right mutations, but that process is random and dangerous.

A Crawl character’s class provides a further development of items and skills, sometimes with a couple of extra perks. Class plays a much greater role in starting skill allotment (but not aptitude) than role, but both factor into it. The following fact is so important to understanding Crawl’s class system that it gets a paragraph to itself, and in boldface too, just to impress its importance:

Class provides no benefits to a character beyond startup; once the game begins, everything a character follows from his actions.

Just because you’re a “conjurer” does not mean you have to cast even a single spell. You could pick up a knife and start training combat skills, and win the game with them, and the game will not make it any harder for a character to do this other than the difficulty he will have starting out with no skill in that area, and whatever stat deficiencies he might have, which are due more to race than class. Likewise, a character beginning as a “Fighter” can pick up the basics of spellcasting from reading random scrolls, then find a starter’s spellbook in the dungeon and begin using magic, provided he is willing to go with light armor and has a good-enough Intelligence stat. It would not be an easy road, but it is possible.

dcss2.pngThose stats are also quite important. Dungeon Crawl has only three: Strength, Dexterity and Intelligence. Their improvement is mostly tied to experience level. At periodic levels one randomly-chosen stat (which is picked depends on character race) is increased by a point. Additionally, every three levels the game asks the player for a statistic to improve. Strength influences combat damage and skill using some weapons. Dexterity plays a role in combat hit chances, attack evasion and skill using the rest of the weapons. Intelligence is needed to memorize spells, and to cast them more successfully once memorized.

That leaves deity worshipped. Most, but not all, Crawl characters begin the game without a religion, and can choose to remain unaligned in that way. (Characters of the Demigod race, in exchange for markedly higher stats and stat growth, are forbidden from joining a religion.) In a special dungeon branch called the Ecumenical Temple some way down, a character can pick out one of twelve gods to worship, each providing special gameplay advantages in exchange for following its dictates. Once a religion is joined, the player can leave at any time by using the Renounce Religion ability, but depending on how much time the player has spent in that religion and which god (if any) he changes to, doing this will incur the wrath of his former deity.

All of these aspects combine to provide a very deep set of gameplay possibilities for each combination of race and class.

How Crawl steers characters along their path

A fact that is not obvious to the Dungeon Crawl newbie is that all of the character races and classes have the same skills. When you check your skills at the start of the game with the ‘m’ key, you will only see a few options available. Despite this, all skills are possessed by the player character. The skills not displayed are just considered to be at “level zero,” and progress towards advancement in them is still tracked, just not displayed. These skills are also called untrained. Whenever the player does something that involves an untrained skill, a couple of points will be diverted to it. Since one of the functions of having levels in a skill is an increase in the number of pool points that can be diverted to it in from a single practice, even with a full experience pool it may still be awhile before the skill advances as points get diverted to the player’s higher-level abilities. (A foresighted player can help prevent this by disabling skills from the 'm' screen, which doesn't actually turn them off, it just makes the game much less likely to assign pool experience to them.)

So every character class can learn every skill, and no class uses special rules to provide its gameplay features. How, then, does Crawl avoid the problem of classes being too similar to each other, such as with most of Nethack’s classes?

Starting out is difficult without a “gimmick,” a system by which a character can prosper in the early going.

A character with no skill in anything would be quite hard to play. The numbers are stacked against such a character; starting stats and skills tend to be just enough that most players will need to rely on class-specific abilities to get a leg-up on the monsters. These skills make possible the character’s engine, the system by which he kills monsters and earns experience within acceptable levels of risk. Having an engine is not strictly necessary, and for some race/class combinations is as simple as walking up to monsters and hitting them with stuff, but without one the player will have to resort to making use of random items and extreme tactics more often, strategies that bring with them necessary dangers.

The hit-them-with-stuff “gimmick” is more conventionally termed melee speciality. Spellcasters that start out with conjurations can also survive in a straight-forward kind of way. But then there are centaurs, which are very fast and great with missile weapons, but must eat more often than most other races, and so must keep on the move more. And then there’s Spriggans, who are extremely weak physically but naturally stealthy, and so they can make a good living off of stabbing sleeping monsters before they wake up. A Spriggan Enchanter (about which more will be said before long) gets a spell that can often put a single monster to sleep long enough to get in a single, almost certain-to-hit, ultra-high-damage critical strike that can wipe out even some very strong monsters instantly.

dcss3.pngMany of the races are built off of these kinds of tradeoffs. The more extreme the tradeoff, the more the player will have to deviate from basic walk-up-and-smack-em play to survive and prosper. This can change the game a lot for some classes, restricting some basic abilities and/or making interesting new ones available. In a way, Crawl is actually a variety of different roguelike games that happen to take place in the same dungeon; the path for success for a High Elf Wizard is very different than that for a Minotaur Bezerker, but they are both quite alien compared to a Vampire Anything.

Beyond race, it must be remembered that Crawl’s classes are, for the most part*, identical after the game begins. Chances for advancement are the same for a human Fighter, Wizard, Assassin, Transmuter or Wanderer. What matters is how they make use of the skills they begin with, and how well they can diversify away from it, which usually relies on item generation.

* What do I mean by “for the most part?” Some classes start out with a god that other classes cannot pick up until they find an altar, usually in the Ecumenical Temple found between dungeon levels 4 and 7. This is a fairly major advantage for those classes, but it mostly just gives them a good start. Additionally, elemental magic skills, including any levels possessed at the beginning of the game, make other elemental skills harder to learn. These differences may delay character advancement in some skills, but the player can always overcome them with more practice... provided he lives that long.

None of this would matter for much if Crawl were not a hard game. The game is very finely balanced, with the knife’s edge very close to the limits of most characters’ abilities at the start of play. Further, the player does not have the luxury to do much grinding for skills due to the food system which, while not as hard as Rogue’s, does limit the amount of food that can be reliably found in a game. It is useful to think of character advancement in terms of experience gained vs. food consumed, with more efficient improvement techniques providing for a potentially more-powerful character later on.

If food were not a resource of hard scarcity, then the player could use the monster generation of early levels as a way to gain experience points, and those points could then be put into skills through practice a lot more easily. So it could be said that Crawl’s food system drives its design. Even the small number of Crawl races that do not need to eat have some mechanism that forces the player to search for resources. The need to find more treasure is what makes the opportunities for experience and skill gain meaningful.

That is enough for this time. Next time we’ll take a look at some of those skills and whatever dungeon-related itches they scratch.

Roguelike news:

It's a little later than expected, but Keith Burgun informs us that his company’s iPhone roguelike, 100 Rogues, has entered closed beta. We interviewed him about that some months ago if you’ll remember. He also mentions that they're having an open beta before too long.

An interesting new roguelike, just over a month old of this writing, is Brian Walker’s BRogue.

In Japan, the awesome videogame TV show Game Center CX, in which a middle-aged man tries to complete old video games, recently tackled Shiren the Wanderer! It was hard enough that they did it over two episodes. No English translation of those episodes yet exists, but episode descriptions are available at Crunk Games. Episode 1 - Episode 2

While I find the show interesting, we have noticed a couple of problems with the playing of this one, notably, just randomly getting swords up to +11, before finding out about the Blacksmith? Seems a little questionable (I have similar problems with the episode about Solomon's Key honestly), but even with some admitted cheating (copying save files) and assistants farming items it takes poor Arino 26 hours of playing to beat the Tainted Insect. Now let's see him do the Final Puzzle!