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

I've said this about first-person shooters before, but in the interest of fairness, I admit it's completely appropriate to say it of roguelikes too: from one quite valid point-of-view, they're all the same game.

Of course when you look at the games up close this assertion falls apart. This perspective completely discounts Crawl's razor-sharp play balance, ADOM's surprising expanse, Angband's epic struggle against the odds, and Nethack's amazingly complex, interlocking gameplay features. But the core of what makes roguelike games was invented back in Rogue, and a big part of that is the item system practically all the games share.

Many of these items are randomly scrambled when a game begins. If the player saves his game (thus ending his session) and loads it back in later (which erases the save), items will retain their identities. Purple potions will still do whatever they did before the save. But if a new game is started, it will have re-randomized items, and if the player dies, all the item identities figured out are completely lost.

Here is an overview of the primary item categories, with an eye towards a closer examination of each in the future. (Afterwards, we'll have a brief recap of the winners of the 2008 devnull Nethack tournament.)

roguefood49.png
One-use items, when used, are gone. There is nothing random about them. In most roguelikes, the basic food ration item is like this, a simple, reliable object that fills the player's stomach. Most one-use items like this are types of food.

Since these items are not random, their purpose in the game is more strictly resource-management than the more unpredictable things left by the dungeon litterers. But what happens if there's simply not enough food to survive? Rogue handled this by scattering a few guaranteed rations around its levels, Nethack, ADOM and Crawl players can eat monster corpses, and Angband contains a nearly guaranteed source of food in town.

But the lesson here is one that a lot of people who make commercial roguelikes could stand learning, that non-random items aren't that interesting. Basic roguelike play involves completely searching each level, and if the items are all known then their uses, unless it's loaded with Rogue-and-Nethack-style alternate purposes, will be obvious. Ultimately, roguelike play is about the items found on the ground versus the monsters found in the rooms and halls, with the player serving as the role of intermediary. Anything that detracts from, or makes obvious, those decisions the player must make weakens the game. If all the items found on the ground are pre-identified, like food rations, then that's correspondingly less the player must do to affect his chances of survival.

roguepotion49.png
One-use random magic items are, most often, potions and scrolls. These items are used up when consumed. This plays a major role in the identification game, for even if the result from using the item left no doubt as to its purpose, the item is still lost. Because of this, these items tend to be fairly easy to figure out; the primary cost to ID-by-trial is item loss and negative effects from bad items. Some of these might be bad indeed: the most dangerous item in Rogue was the potion of blindness, which made the game nearly unplayable for a few hundred turns.

A few roguelikes, in addition to non-random food rations, offer one or more random food types. Berries and mushrooms, natural sources of sustenance that may have unexpected, even dangerous, properties are common choices for these. (Shiren's herbs are also like this, even though their other aspects make them a closer analogue for potions.) They are often used as a kind of variant type of potion, and so can, indeed must, have powerful effects, for the nutrition from eating one is almost never worth the risk of identifying it unless the player's character is in starvation mode.


roguescroll49.pngWearable random magic items are usually rings and amulets. Rogue itself only had rings, which are pretty widely-spread. (It also had the Amulet of Yendor, but that's most of a goal item, see below.) Hack added random amulets, but not many other games have gone along with it. These items are not generally lost when used; they continue to perform their function over time. This eliminates one of the major costs to ID-by-trial, so to compensate most of their functions are subtle. No normal player is going to notice doing an extra +1 to damage. Also, since items with bad effects could have their effect nullified just by removing it, these items are often cursed, meaning the player can't remove it once it is worn without using special means, usually a specific type of scroll.

(Some more recent roguelikes, especially Hack, ADOM and Shiren, have generalized the idea of curses to cover all kinds of items. There, a cursed item will work badly, or in a frustrating or dangerous way, or simply not at all. Some of these games also apply the idea of the opposite of a curse, that is, a "blessing," which similarly improves an item's functioning.)

In D&D, and some other CRPGs as well, a curse is a specific type of bad effect that may or may not make it hard to stop using the item. In roguelikes, a curse is a property of the item separate from its purpose. The curse can be lifted and the item used normally. This doesn't mean the item is suddenly good for you, just that it can be worn and removed freely.

There is one further cost normally associated with these items, and that is food consumption. It's another of those obscure roguelike rules that wearing a ring makes you slightly hungrier over time. (And a couple of rings, hunger and regeneration, make you much hungrier.) The effect this has on the game varies by how perilous the player's food needs already are; in Rogue, the player is often only a ration or two away from starvation, while after the early game in Nethack players usually pass up far more food than they eat. But the presence of the food drain makes clear that developers believe such long-lasting benefits should be compensated for, even if they're relatively minor.

roguewand49.pngRandom multi-use items are a mixture between the two previous types. Nearly all roguelikes use wands (or sometimes, sticks) for this, but there are other types. Usually they have a limited number of charges, and when the item runs out it's useless.

Wands are even easier to figure out than potions and scrolls, and probably still have several charges remaining after an initial trial use, so the resource cost for testing them out is much less. And since they're not wearable, curses don't apply in many roguelikes (unless it's a game with generalized curses). So to compensate, the risk of bad effects is rather greater with wands than the other types. This is helped in that nearly all wand functions involve doing something to a monster. Testing one on a weak monster could be a fatal mistake if the wand turns out to be polymorph, but even something like fire that could whizz by the foe, bounce off a wall, and hit you.

Nethack's spellbooks are a rare example of a multi-use random item that doesn't have charges, or at least not charges that substantively affect the game. (As with many things in Nethack, the precise truth would take a long time to explain.) They must be identified, which might be done by trial. Notably, however, other than the difficulty level of the spell all the bad effects from failing to learn a spell, or reading a cursed book, are the same across spells. Once a spellbook is read, it's always known. (I've always considered this as kind of a missed opportunity; why not make the player figure out what a learned spell does?)

rogueweapon49.pngThe place of normal equipment is easy to overlook when explicating the roguelike item system. Neither weapons nor armor in these games are randomized in the same way as the other types. A dagger will tend to operate like a dagger, even across games. Leather armor retains its rustproof property without having to discover it each time. The player can acquire knowledge of which items are better and worse and rely upon it between games.

(An aside.... When there is a lot of this knowledge, like of rustproof leather, to acquire, it can contribute to a feeling that the player must "pay his dues" building miscellaneous knowledge before he's allowed to do well at the game. In the old days, this information was either regarded as common—lots of early roguelike players knew about weapons and armor from playing Dungeons & Dragons—or a just reward for clever deduction; now, people see it as just another thing to look up in a FAQ. This is one reason that I believe the future of roguelike gaming leads in the direction of more randomized game elements, as it makes a game more resistant to spoilers.)

While the basic function of normal pieces of equipment are the same, they are still, in a small way, randomized. Weapon and armor pieces carry a property called a plus, named after the concept from Dungeons & Dragons. The plus of an equipment item is hidden from the player until it is identified. The conventions with different games vary; Rogue and Hack-likes reveal the plus of armor as soon as it is worn. Unlike with random magic items, once one +3 long sword is known others of that type are not instantly discovered. Each individual item must be identified to find its plus. Equipment that carries a negative plus (a.k.a. minus) may also be cursed, another property that must be discovered on a per-item basis.

Some games introduce other kinds of per-instance item properties. Angband and ADOM's ego weapons follow this property, an attribute that can be tied to an object that must be identified to discover. Many roguelike games also have artifact items to find; usually these types are identified to the player ahead of time (they are obvious information due to the object's status in legend), but ADOM in particular treats artifact status and name as information that must be identified. But IDing one ego weapon doesn't give insight into others of the same type, and knowing how to make best use of an artifact tends to be spoiler material.

roguearmor49.pngRandom magic equipment is something a few games introduced in order to approach Dungeons & Dragons to a closer degree. Mostly, these games are the Hack-likes and ADOM. They include types of weapons and armor that have generic base properties, but also a random magical function. The base properties are tied with the description (snow boots), but each item also has a "real" type (speed boots). The tie between the appearance and magic power is randomized in the same manner as the other magic stuff, so these items vary from game to game. Jungle boots in one game might be boots of water-walking, but in another might be fumble boots.

In a few cases, the "description" of one of these items implies another use, and Nethack's Dev Team is canny enough to also build those uses into the game. Riding boots and gloves are random types, but regardless of their magic, they additionally make it easier to mount and ride steeds. Snow boots make it easier to get around on icy surfaces, although it's possible that its magic type will turn out to be fumble boots, hardly an improvement.

Interestingly, although magic rings have been around since Rogue, Nethack and ADOM provide for random magic armor pieces, and both those games plus Crawl and Angband have artifact weapons (those other two adding "randarts" whose properties vary between games), no roguelike game I'm aware of, major or minor, provides for random magical weapons, on the same scheme as potions, wands and rings.

Finally, there are the goal items. Roguelike games are more about the dungeon than the monsters, so more often the goal is to get some object and escape with it more than to Kill Foozle. This gets a category to itself because, in play terms, these items are often useless, or have an obscure, or purely winning-oriented, kind of utility. Rogue's Amulet of Yendor confers only two abilities: much reduced hunger, and the ability to go up stairs, both of these things being necessary to win. Nethack's Amulet major function is to prevent level teleport, which is necessary to stop the player from escaping thirty seconds after he finds it. Nethack's three "key" items, needed to get the Amulet, also fall into this category. Dungeon Crawl's goal items are runes of which an operator-adjustable number are needed to win (commonly 3, but some players try to collect 13 or even more), and the Orb of Zot, which, like Nethack's Amulet of Yendor, exists primarily to make the player's life more miserable while it's carried.


Surprisingly, marvin did not take away top prize again in this year's devnull Nethack tournament. I sometimes make jokes about his awe-inspiring Nethack skills, but he is an amazing player. This year though he still won 11 games, certainly no slouch. marvin's position as the maintainer of the official Atari window port of Nethack means that his slightly-diminished performance this year gives one pause. Could it be that the reason is a secret round of DevTeam testing aimed at a new release? Interested parties are invited to speculate upon this possibility on message boards throughout the internet. (The preceding statement is an unfounded rumor with no basis in reality. It represents wishful thinking only, and in no way represents a sneaky tactic to prod the Dev'vers into releasing a new version.)

There were more players vying for the trophies this year than any before, likely because of Slashdot's gala 10-year-anniversary article on the tournament. Best of 13, which goes to the player who wins the most games in a row of different alignments, races and roles, was won by hillance. Most ascensions (18 in the 30-day tournament period) was hard-earned by adeon.

The fastest victorious game, in game-time, was played by tenaya, coming in at 8,023 turns. The fastest in real time was played by adeon, bathed in radiance after only in two hours and 22 minutes. The lowest-scored victory was won by nuslayer, who had a scant 24,920 points when he went to his reward.

theta once again won the most types of death trophy, this time "discovering" 161 of the game's reasons for cessation of operation. This was less than last year, but still an impressive display of mortality.

This year's special challenge turned out to be an entire other roguelike game, ZAPM! A science-fiction game inspired by Nethack, it got its own high score, winner and most causes of death scoreboards in addition to its overall challenge board. The game looks interesting in its own right, and may well get its own @Play focusing on it before long. In any event, 87 players attempted it, and 34 completed it.