I want to write today about leveling – a characteristic of roguelike games that is shared with most MMO’s and tabletop roleplaying games. It’s one I’ve recently implemented support for in my own game program, so I’ve recently done some thinking about what it is and what game design purposes it serves.

Characters in these genres “level up” every so often, usually in response to “taking experience”, which in turn usually means killing things and taking their stuff. In some ways this forms the very basic cycle of gameplay in these games, and as such it’s one I want to examine. So I’m considering a few cases that occur in games I happen to know.

My introduction to roguelike games was Angband. And Angband implements a pretty simple experience ladder. Players range from level 1 to level 50, and each level takes about a half-hour to one hour of gameplay time to get if you’re pushing as hard as I know how to play. I’m sure there are players who can do it much faster and players who take much longer – this is just my experience of it. A starting character has about ten to twenty hitpoints depending on race, class, and starting constitution. Constitution is maximized (and therefore no longer a difference between characters) by the early midgame, and by the time a character wins the game they have usually achieved maximum level, and have about nine hundred to eighteen hundred hitpoints depending on race and class. So this is almost exactly an increase of a factor of 100 in the character’s ability to take damage, and most other skills and abilities are balanced to that increase.

But Nethack has a distinctly different rate of power inflation. Nethack has characters start with usually ten to twenty hitpoints, which is similar to Angband, but winning nethack characters (as played without taking advantage of spoilers and metagames) have about two hundred fifty to four hundred. Most of the other abilities a character has and threats a character faces are balanced to that scale of increase. In contrast to Angband’s factor of a hundred in power difference between starting and ending characters, this is a factor of only about twenty. Or if you prefer to think in exponential terms, it’s only a bit over four doublings in toughness as opposed to nearly seven.

Another distinction between Angband and Nethack is that the style of play I specified above (not taking advantage of spoilers and metagames) is actually fairly rare. Characters played by that style are nowhere near the top of nethack’s brutal experience scale; it simply isn’t possible to get to the top of that scale in normal play. But about half the winning characters have substantially more than the maximum number of hitpoints they could possibly have gained just by reaching whatever level they have, and about a quarter have actually reached the top level of nethack’s starkly impossible experience scale.

This is due to the availability of meta-games involving healing potions, wraiths, vampires, succubi, nurses, polymorph spells, the ability under some circumstances to eat certain magic items normally worn or wielded, the brutally stupid idea that certain spells might be better to cast from cursed scrolls or while confused, and other inside information. The vast majority of players who take advantage of these meta-games have not learned about them in the context of actually playing the game, which makes them “spoilers”.

Spoilers are a tangential topic here, and while we need to be aware of them to evaluate the effects of power leveling, this is the article about the effect on gameplay of different scales of progression in power levels. Another article will have to be written about the more direct effect on gameplay of spoilers.

To get some notion of the purpose of the leveling mechanic, let’s have a look at other game types where the player controls a character. In sidescrollers, there’s no powerleveling mechanic. The character simply moves to the goal as fast as possible. Navigating the barriers is what the game’s all about. Interactive Fiction games, as a rule, don’t use power leveling either; your character, if you win the game, ends with about the same capabilities he started with. In an Interactive Fiction game, you typically start with some objective, which I’ll call A, then you discover that you need to get three or four other things, which I’ll call B, C, D, and E, done in order to achieve that; then you discover that to get B done you must first get three or four other things done, which I’ll call F, G, and H, or you must do some harder thing, which I’ll call I. So you do F, G, and H, then discover that you have to do I in order to get C done, but you can’t do I unless you undo H; so you undo H, do I instead, and get B accomplished the second way. Then you look to see what else you have to do in order to get C done…. and so on in an interlocking series of puzzles and feats that, when you eventually unravel them all, brings you to the conclusion of the game.

Now what is actually happening here? In the sidescroller, the game is about getting through the map as fast as you can, and in the course of getting through the map you interact with all the content of the game. And that style of interactive-fiction game is putting a very large, very precise set of requirements on the player, which must be satisfied in very specific ways. These requirements also ensure that you interact with every part of the game’s content, because you have to coax that rat out of that hole using that peanut butter in order to get the king back his crown. Or you have to have that pocket lint in the pocket of your bathrobe at the end of the game in order to head off the alien invasion fleet. Or you have to find the grocery list sticking out from under that carpet in order to prevent the mad doctor who is controlled by a sentient alien meteorite from sucking out your girlfriend’s brain. Or whatever.

I’ll probably have to revisit interactive-fiction as a genre when I do that article on spoilers, now that I think of it, because some of these puzzles failed as puzzles; there simply wasn’t sufficient information available in the context of playing the game to learn what you needed to know to solve them. Spoilers, whether from other players or from the company help line, were the only way to get past some barriers.

Now consider roguelike games again. If you were as powerful at the beginning of the game as you were ever going to be, the optimum strategy would be to play it like a sidescroller; go as directly as possible to the end boss while avoiding engaging with all intermediate threats, and defeating the end boss to win the game. What is the reason you’re interacting with all that midgame content? The reason is to invoke experience leveling and improve your equipment kit. In other words, the power leveling mechanic is taking the place of all of that intricate interlocking puzzle from the interactive-fiction games, in preventing you from winning unless you spend time interacting with the game content. And it also allows a more “open” map than a sidescroller, in which the player can choose his own route, because the map doesn’t have to constrain the character to interact with everything along the way.

The leveling mechanic in roguelikes replaces all those things that you were collecting or feats that you were accomplishing in interactive-fiction games, and also replaces the direct barriers to navigation in sidescrollers. It is whatever you need to move on to the next part of the game, and the generality of “experience” ensures that you can get it from any of many sources. Whatever you take on, what you get from it will be exactly what you need to access the next part of the game’s content.

Power leveling even compensates, to some extent, for unexpectedly difficult content or for players with less skill. If the game throws out a challenge that a player cannot overcome, leveling means that additional means to overcome it will be available later. It also means skipping that challenge is a viable strategy, because if the player’s midgame objective is simply to increase his power level, there’s nothing absolutely required about defeating that particular content; there are other ways to power level.

So it becomes a natural part of roguelike design because roguelikes generate content procedurally; In the first place, it would be much harder to generate the kind of interlocking puzzle that an interactive-fiction game relies on by procedural content generation. In the second, it allows the game to continue even though the randomness of procedurally generated content occasionally presents something that the player or the player’s character cannot immediately defeat.

And finally, if that’s a correct understanding of its purpose, we have at least a few guidelines about how much it is good for game design in the roguelike genre, which is more or less defined by its absolute reliance on procedural content generation.

At one end of the scale, we can imagine a roguelike game with minimal or no power leveling mechanic. It would be in the best interest of the player to bypass every possible risky engagement and proceed as rapidly as possible. As a game designer you’d have to address this by generating incentives, puzzles and requirements along the way that required or enticed the character to engage the content – or at least navigation barriers that prevent the character from being successful in his or her attempts to bypass all of it. In order to make these requirements along the way meaningful, you’d have to make the levels persistent, or the player could simply skip them by forcing the level where such a challenge exists to be generated again. And once you have persistent levels, you cannot, even once, generate a challenge that the player can neither defeat nor bypass, or you are making a game that absolutely cannot be won because there’s no way to come back to that challenge with a higher power level. It would be a hard game to write, but might be a good game to play.

On the other end of the scale, we can imagine a roguelike with a leveling mechanic even more extreme than Angband’s. The player must spend a long time engaged with threats of each level before threats of higher levels become possible to engage. If the player proceeds to deeper levels before leveling enough, the character will get stomped like a bug. There is no need for puzzles or barriers along the way, because engaging every part of the content (or at least a lot of content at every level) is necessary to level in order to make the endgame possible. Non-persistent levels become entirely reasonable in terms of gameplay. Without larger-scale barriers to be overcome or puzzles to solve, there is no need for persistent levels to ensure that players don’t bypass those barriers or puzzles.

And once levels are non-persistent, unlimited content can be generated. This enables players to succeed with different playing styles, or compensate to some extent for greater or lesser skill, by more or less aggressive diving behavior. Non-persistent content allows less aggressive (or less skilled) players to engage more content at each level, while maintaining a higher power level relative to the content they’re engaging. With persistent levels, the amount of content that exists at each level of challenge is limited.

So, the difference in approaches to leveling are part of different strategies in roguelike game design. They go together with the subgoals and puzzles of Nethack versus the complete absence of such things in Angband. They also go together with the persistent levels in Nethack and non-persistent levels in Angband. And they explain how it’s possible that Nethack characters of less than the maximum possible level – sometimes far less – occasionally win the game, even without invoking the specific set of metagames that can raise their level or hitpoints or both far beyond what is possible from normal play; with less of a power gradient and more variation in character capabilities from other sources, differences in player skill and player knowledge can drive a much larger difference in whether the endgame is accessible.

So what’s better game design? As with so many things, it comes down mainly to a question of what you’re trying to do. And having implemented the infrastructure to support player experience levels, and thought hard about it, and having written this article, I’m still thinking about it. On the one hand, I feel that having the player trying to accomplish specific subgame objectives rather than just killing everything for experience and loot makes a more interesting game, and that argues for keeping the power leveling to a minimum. On the other hand, procedural content generation of content that will keep a player engaged without power leveling is harder to implement. A certain amount of power progression provides some satisfaction for the player, some regulation about which game content is available when, and some flexibility about how players with different amounts of immediate skill or knowledge can find a style of play that makes most content available to them at some point.

And on the third hand, the more we eliminate power leveling in favor of subgoals and puzzles, the more we produce a game that favors metagaming – players using knowledge they didn’t acquire in the context of actually playing the game, which effectively prevents new players and players who haven’t sought out spoilers from ever winning, and that isn’t good game design either.

I feel that puzzles and subgoals are good game design compared to leveling, but good puzzles and subgoals rely for their solution only on information that most players can and will acquire in the course of actually playing the game. When the player cannot acquire the necessary information, you get metagames instead, and avoiding them is very hard in a game where you’re procedurally generating as opposed to writing the content. As all interactive-fiction gamers know, it is even hard when some rational human being is actually writing the content. Metagaming – in which the game is unfair in favor of spoiled players – I hate passionately, and will very happily accept as much power leveling as necessary in order to avoid.

I also feel that a certain, fairly small, amount of power leveling is beneficial in that it helps some players stay engaged and enjoy the game – but that if there is so much power leveling that players take to ‘grinding’ because they believe that it is the only approach that will allow them to reach the endgame, then I have failed the player because the way to play the game and have fun is not also the way to win.

So, I guess that’s my conclusion. Power leveling in roguelike games is an element of game design that I will need to use in order to avoid creating a blatantly unfair spoiler-driven game, and which I should keep to a minimum in order to avoid creating a blatantly unfun grinding-driven game.

I hold out some hope that I can have both less power leveling and less spoiler influence than in nethack, leaving the difference up to player skill as opposed to spoiling. But in order to do that I would need to have to bring some new elements to roguelike procedurally generated content, so that subgames and puzzles that are meaningful in the context of the game can be generated by them. These elements are factions, nations, clans, etc, with goals and desires of their own and modeled loyalties, enmities, and other relationships.