Tales of the Rampant Coyote: Exception Driven Game Play

Rampant Coyote has an excellent post on exception driven game play in RPGs which ties in rather nicely with my series on randomness in RPGs(and not for the first time). It's a must read if you often find modern PC RPGs lacking compared to their less refined ancestors.

Tales of the Rampant Coyote: RPG Design: That Which Is Not Forbidden...

Very interesting post at Tales of the Rampant Coyote which ties in directly with my concept of player input variance and open-endedness in RPGs. In discussing two of Gary Gygax's published pen-and-paper adventures:

But both Tomb of Horrors and Necropolis left a lot up to interpretation by the Dungeon Master (the person who "runs" the game). And I try and run my games by a guiding rule which, lamentably, tends to be ignored in more recent editions of the game, and ignored by players who are used to computer games: That which is not expressly forbidden is fair game to try.

And the conclusion:

The problem is that - for the most part - RPGs aren't made as anything resembling simulations. That's too difficult, and it is too hard to put the player on the kinds of rails that many designers prefer. So spells have very particular, extremely limited uses, and tend to be more of the "blow crap up" variety. Spells that provide knowledge, hints, or "intelligence" are subject to exploit in single-player games, as the information they provide to the player is persistent, even when the player reloads the game immediately to 'restore' the expended spell.

Our worlds are just too restrictive to allow this kind of play. But do they have to be?

They definitely do not have to be so restrictive. One way to deal with the issue is through random procedural content. Another is by limiting the ability to save/reload. Roguelike developers already care about these issues. I'm just waiting for the mainstream to catch on and it makes me very happy to see someone else talking about it.

Randomness and the Black Swan in RPGS(Part 2)

In the last two entries of this series I discussed the role of randomness in RPG task resolution mechanics. I also demonstrated the current shift toward increasingly deterministic resolution systems in electronic RPGs, and I explained why I felt this was a bad thing. Now it is time to don the mantle of chaos once again and discuss the other major set of RPG systems influenced by randomness, content generation systems. I will also introduce a couple of conceptual tools which may prove useful in my analysis.

In most P&P RPGs, randomness exerts its influence from the very beginning, at character creation. Anyone who has played any edition of Dungeons and Dragons is familiar with summing the results of 3 6-sided dice to determine a character's ability scores like strength, intelligence, etc. Hit points for both players and monsters are also determined by rolling dice. These are all simple ways in which randomness feeds into the content generation systems of RPGs. Simple randomness ensures that all orcs aren't exactly the same. The inclusion of this type of randomness also means that one player's 3rd level fighter may be more powerful than another player's. However, the greater the number of attributes with randomly determined parameters, the more likely most characters will trend towards average overall and the more likely most characters will excel in some area.

Random character generation was one of the first random elements of P&P RPGs to be removed from electronic RPGs. Aside from licensed D&D games and roguelikes, few electronic RPGs still have random elements in their character generation systems. In many electronic RPGs there isn't even such a thing as character generation. This is entirely logical for story-based games and games without much in the way of randomized content generation. But, when the challenges to be met aren't predetermined, why shouldn't an RPG utilize randomness from the very beginning to vary the game experience?

But when most people think of random content generation, they think of randomly generated environments, encounters, and rewards. All of these elements were present in D&D in the 1970s. In general, every encounter area was expected to have an associated wandering monster table. The DM made periodic wandering monster checks, and if one occurred, the appropriate table was consulted. This allowed the DM to generate encounters on the fly while the party was travelling long distances or camping. Wandering monster checks could also be used to punish players for making foolish decisions like making a lot of noise in a dungeon. The AD&D Dungeon Masters' Guide included encounter tables for all types of wilderness terrain and dungeon levels. There were also tables to generate treasure on the fly for these and other encounters. Each monster type had an associated treasure type with an associated set of tables for randomly determining what treasure a monster of that type might have in its possession. There was a very slight chance that even a weak creature could have a powerful magical item.

With the release of the AD&D Dungeon Masters' Guide in 1978, random content generation was taken to the next level with the inclusion of a complete system for randomly generating dungeons. Level layouts, environmental details, tricks, traps, treasure, and enemy encounters could all be generated on the fly by the DM. The inclusion of such a system says volumes about the way at least some players played P&P RPGs. It is also some indication that, at least for some players, the mechanics of D&D were fun in and of themselves.

Of course, it should be mentioned that the Dungeon Master's Guide explicitly states that the content generation charts contained therein are not to be taken as Law. The DM can freely ignore any result he feels will unbalance the game or which he simply doesn't like. However, potentially unbalancing random events are ameliorated somewhat by the flexibility of a game played between humans. If the players suffer a stroke of bad luck and encounter a high level wandering monster they have a practically infinite number of options available to them, limited only by their creativity. They can scatter in all directions, they can describe in detail their attempts to hide or escape, etc. The DM can then reward quick thinking and good ideas. Contrast this with early electronic RPGS where the only option besides fighting was selecting RUN from a battle menu and it is easy to see why the range of random encounters might have been restricted.

Now I would like to introduce two concepts: content generation variance and player input variance, hereafter referred to as CGV and PIV. CGV is a measure for the range of output possible by the content generation system given a fixed set of parameters. CGV is easy to quantify in a video game because it is essentially an algorithmic function with specified parameters. PIV is harder to quantify because it always involves human input. Sometimes, such as with menu systems, the human input is very restricted, but many modern RPGs are much more open ended than that. In addition, PIV is also a result of the way in which human input is processed and translated into game actions. PIV takes into account randomness within the game's task resolution systems(see Part 1 for more details), for instance. So, PIV is defined as the range of results from all possible player actions in a given game situation. PIV will remain a fuzzy concept, but will suffice for our purposes. A P&P RPG played between humans will be considered as having maximum values for both CGV and PIV.

Next time I will continue the discussion of CGV and PIV. I will discuss how they relate to one another and how they have changed over the years in RPGs.

Randomness and the Black Swan in RPGS(Part 1.5)

Before looking at the random generation of game content in Part 2, I want to elaborate on a point made in Part 1 concerning combat mechanics: "The point of this comparison is that predictability isn't necessarily more fun because it is "balanced", sometimes it is just boring. And if it would be boring in a pen and paper game, then fundamentally, it is probably still boring electronically. It is just dressed up in a such a way that people put up with it."

Admittedly, I glossed over all the elements which can compensate for predictability in a electronic RPG's combat system. For one, the game can require the player to manage an assortment of abilities/skills in a real-time situation where tactical decisions must be made quickly and the player has to adapt to evolving situations. This is the approach taken by most MMORPGs, and some, like World of Warcraft, do it well. I still argue that too much predictability is a negative, though. Few encounters in such a system are unique and there is a very clear delineation between encounters the player can handle and encounters he can't. It is common for a player to find himself in a situation where if he draws/lures one enemy he will win fairly easily, but if he draws two he will die. This type of certainty reduces the depth of decision making. The challenge is moved firmly into the realm of player execution.

Interestingly, Final Fantasy 12, which I used as an example of predictable combat, seems to recognize that its base combat mechanics aren't very interesting on their own. Recognizing that the opportunities for complex decision-making in Final Fantasy combat are fairly rare, the developers automated the system. The combat in Final Fantasy 12 plays out in real-time according to simple, player-assigned AI rules controlling each character's actions. The player can pause the action and intervene if necessary, but this is rarely required outside of boss battles. Instead, the player takes satisfaction from tweaking his party's AI until it is as self-sufficient as possible.

The lead developer of Battle for Wesnoth, an open source turn-based strategy game, explains why his game has a substantial random element in its combat mechanics here. He makes a number of excellent points that apply just as well to role-playing games:

In Wesnoth we want a player to plan out a complex situation, to estimate carefully the possibilities. To have to work out a good strategy. If a player can simply rely on all sorts of assurances that their units will hit, they don't have to do any of this. Sure, there will be a certain amount of fun to planning out a situation where you can set up a cool 'domino effect' of enemy units going down as you attack them. But this is nothing to do with the skill of planning out a real strategy in a dynamic situation where you have to consider all kinds of contingencies.

...

Trying to "slide down the scale" of luck would simply make Wesnoth less interesting, in my view. Suddenly the difference between a two attack unit and a four attack unit would be trifling instead of critical. Not near as much planning would be necessary. Simply, in my view, Wesnoth would become less fun.

The point I want to make in this post is that although predictability can be compensated for, especially if the combat system plays out in real-time, an infusion of unpredictability into the system can make it even more interesting.

Randomness and the Black Swan in RPGs (Part 1)

Raph Koster wrote an interesting post on the connection between games and the ludic fallacy as discussed in Nassim Nicholas Taleb's bestselling book The Black Swan. For those unfamiliar with the book, a black swan is an unpredictable but highly impactful event. Raph ends his post with a question:

"Rather than just bemoan this, I'll instead issue the challenge: what is the fun game that features black swans, phase transitions, and the catastrophic 100-year flood? How do you sculpt a system that does this without chasing away the newbies?"

This is an interesting question, and one which has a simple answer if not constrained to electronic games: pen and paper roleplaying games. The presence of a human game master lends unpredictability, surprises, and ad hoc creativity that electronic games can't hope to match. However, even ignoring the human element, the mechanics of electronic RPGs have been straying ever further away from unpredictability in a misguided pursuit of balance. Since long before I had the Black Swan framework to hang it upon, I have been thinking about the role of randomness in roleplaying games.

Roleplaying games typically draw on randomness for two different major functions. The first is in their action resolution mechanics. The second is in their content generation systems. Both of these got their start in the earliest pen and paper roleplaying games. The earliest electronic RPGs, existing primarily to recreate the tabletop experience, inherited both of these random elements. As electronic RPGs have diverged from their pen and paper roots, the role of randomness has steadily diminished. The results of this process are manifested most clearly in today's MMORPGs and JRPGs. But let's not get ahead of ourselves, even within the early pen and paper tradition, randomness has been afforded varying degrees of importance.

Let us begin by comparing the combat resolution systems of Advanced Dungeons and Dragons and Rolemaster. In AD&D a typical orc has 1d8 hit points and an armor class of 6. A first level fighter with no bonuses will hit an armor class of 6 with a roll of 14 or higher on 1d20(the standard die used for attack rolls in AD&D). At tenth level, the same fighter only needs to roll a 5 or higher to hit the orc. With a standard longsword, the fighter will do 1d8 hit points of damage to the orc if his attack succeeds. So, the first level fighter will hit the orc 35% of the time and it will take anywhere from one to eight successful blows to kill the orc. According to my calculations, the chance the fighter will kill the orc with his first swing is just under twenty percent(0.35 * 0.5625). However, if the fighter had a strength of 18/00(the max a human fighter can have naturally), he would receive a +3 bonus to hit and a +6 bonus to his damage roll. In this case, the fighter would have almost a fifty percent chance to kill the orc on his first attack. Clearly, there is a large random element to combat in AD&D. Bonuses from high ability scores and special abilities and equipment do reduce the relative importance of the random rolls, but these bonuses are usually fairly small. The main constraint to AD&D randomness is that the range of results is clearly defined. If a player's character has 10 hit points and an enemy is attacking him with a longsword, the player knows that his character can't be killed in one blow unless the opponent has some damage bonuses from strength, magic, etc.

In Rolemaster, however, things aren't quite so simple. The attack roll in Rolemaster is an open-ended d100. Normally, results range from 1-100, but on a natural/unmodified roll of 96-100 another d100 roll is made and added to the result of the first roll. This second roll is open-ended as well, so a natural 96-100 will result in a third roll, and so on, ad infinitum. The upper bound of an attack roll in Rolemaster is infinity. Unlike AD&D, the attack roll determines both whether a hit occurred and how much damage was inflicted, with one exception. The final attack roll result, plus the attacker's offensive bonus and minus the defender's defensive bonus, is indexed on a chart for the weapon being used and cross-referenced against the defender's armor type. The result is a number indicating the amount of damage inflicted and possibly a letter indicating a critical hit. Most combat deaths in Rolemaster are either direct or indirect results of critical hits; combatants tend to die in sudden and dramatic fashion and not as much through attrition. There are five degrees of critical hits, A through E, with E being the deadliest. When a critical is indicated, a separate d100 roll is made on the appropriate critical chart. Crticals can result in broken bones, bleeding, blindness, loss of consciousness, and even gruesome death. Even an A critical can kill on a natural roll of 100. Combine the lethality of critical hits with the fact that even the best armor usually can't reduce the probability of a critical hit to zero, and you have a system where any blow can kill. The genius of Rolemaster is that it allows players to manage risk by using any portion of their offensive bonus as a defensive bonus instead. It is rare for players to take a purely offensive stance unless their opponent is stunned.

Now let us draw a comparison between these older pen and paper mechanics and a modern console RPG, Final Fantasy 12. The damage equation for wielding a spear is:

DMG = [ATK x RANDOM(1~1.125) - DEF] x [1 + STR x (Lv+STR)/256]

Notice that the random element ranges from 1.0 to 1.125. Vaan at 50th level, wielding the best weapon in the game(ATK 150) against an enemy with a DEF of 50 yields:

DMG = [150 x RANDOM(1~1.125) - 50] x [1 + 50 x (50+50)/256]

The end result is damage ranging from 2053 to 2438, a fairly narrow range compared to the pen and paper systems. Also notice that successful attacks are the default. Instead, certain equipment like shields provide a chance block an enemy's attack. There is also a small chance to score a critical hit, which results in double damage. As you can clearly see, combat resolution in Final Fantasy 12 is much less random than even AD&D.

To see for yourself how much things have changed since the early days of roleplaying, just compare the combat of a grind heavy MMORPG with the commercial MUD Gemstone IV. Gemstone IV's game mechanics are derived from Rolemaster(in fact, earlier versions of the game licensed the Rolemaster rules and the Shadow World setting). You will not hear Gemstone players using terms like DPS(damage per second) when discussing combat strategy. You often hear complaints from MMORPG players concerning the grind necessary to gain levels, but in fact every MMORPG battle is a microcosm of the level grind. This is especially true of the large raid bosses. You can find a very brief excerpt of a Gemstone IV battle here. I was caught in an invasion of high level plant creatures into my low-level hunting ground and killed. Another player who attempted to rescue me was killed as well. The point of this comparison is that predictability isn't necessarily more fun because it is "balanced", sometimes it is just boring. And if it would be boring in a pen and paper game, then fundamentally, it is probably still boring electronically. It is just dressed up in a such a way that people put up with it.

But even the relatively unpredictable and complex combat mechanics of Rolemaster fall short of the conditions needed for Taleb's black swans. Even without knowing all the bonuses involved, the player can roughly estimate the risks. Looking at the attack charts, the player can see what results are needed to generate critical hits. With transparent mechanics, black swans are by definition impossible because the possibilities are known beforehand. With regard to task resolution mechanics, this isn't necessarily a bad thing. It is the generation of game content which can really benefit from black swans -- like the plant invasion which killed my character in Gemstone IV -- and this shall be the topic of my next post in the series.