HippoPotenuse
  » the skewed views of a large opinion: Persistent World Design
Wednesday, March 31, 2004
  < AI: The Problem >        
Artificial Intelligence is something that is hard to do in any game, in any genre, on any platform. AI in persistent worlds is typically some of the worst. Monsters rarely use their abilities intelligently, and groups of enemies never work together, or demonstrate any behavior other than spamming random attacks, on targets determined by 'hate' lists.

The first primary problem in persistent world AI is pathing. This is quite difficult in any 3d game, and as a result, most games allow monsters to 'clip' through world geometry to charge players. This is intended to alleviate the problems of players abusing the world to 'snag' a mob, and killing it with impunity.

The technical cause of this problem is that pathing through three-dimensional terrain (including collision detection) is computationally expensive. A server simply doesn't have enough processor time to devote to a 'proper' pathing implementation for every mob in its zone. Without proper pathing, it becomes difficult for a mob to recognize when it can and cannot get to a target. Without being able to identify that situation, mobs have no chance at reacting intelligently to it.

Groups of monsters also don't behave like cooperative groups of monsters. Mobs never rely on each other to perform certain roles the way players do. They don't communicate anything (outside of limited Bring-a-friend). Mobs have no sense of 'group'.

Mob tactics are similarly straightforward. Mobs typically track a simple 'hate' list to determine which player to attack. Player actions are tuned to increase 'hate' to provide some semblance of balance. A monster will never realize that it should take out the healer if it wants to have any chance of defeating the tank. There simply is no difference between a healing act or a massively damaging act -- all the mob can know is the severity of the act.

Furthermore, mobs tend to wander around aimlessly, if at all, when not in combat. Encounters aren't challenging, interesting or dynamic. Fighting the blue foozle is essentially the same as fighting the green foozle - regardless of any other circumstances. On the edge of a cliff, in grassy plains, deep within a forest - monsters all simply spam out a series of pseudo-random attacks at targets identified by 'hate' and proximity.

Intelligence in combat is difficult, once again, because intelligence is computationally expensive, and servers don't have many clock cycles to spare. Any added logic for mob processing gets multiplied by the number of mobs in a zone - which simply doesn't leave much room for improvement. Any attempts at intelligence are therefore relegated to a handful of rare mobs.

Mobs don't even have a goal or purpose, much less a plan. Without a purpose or plan, they cannot truly be defeated, let alone victorious. Therefore, the players themselves cannot win or lose. How does someone 'win' an encounter against an opponent who cannot lose?

Generally mobs aren't given a purpose because they are static content, meant to be consumed by the player. Even when mobs aren't a static encounter, designers don't allow them the ability to 'win', because allowing them to win requires that players are allowed to lose. The design consensus seems to be that its better for conflict to have no true resolution, than an undesirable one.

These same problems spill over into the realm of non-combat logic, such as NPC behavior and questing. Goals, tasks, and branching is simplified to keep resource utilization low. So are we hopelessly stuck until hardware frees us from our design shackles? I don't think so. In fact, I believe that the architecture of persistent worlds offers up a natural solution that allows our genre to solve the problem in ways that other platforms and genres simply cannot. Stay tuned for The Big Idea™
 
  < / AI: The Problem >        
Tuesday, March 30, 2004
  < Organizational Downtime: The Suggestions >        
So what can we do to support our community by easing the difficulties of organization?

No Required Classes
If a game is balanced such that nearly any selection of groupmates are an effective force, we can severely curtail the amount of time players sink into finding a group. This is naturally a little trickier than it sounds. For many games, classes are predicated on the concept of specialization and codependence: Mages do massive damage from afar, but need a warrior up front to keep the monsters away; Clerics do healing, so no-one else can heal anywhere near as well as they can.

Removing the 'requirement' of having certain classes, by definition, requires less codependence and less extreme role specialization. There must be a wide selection of classes which can perform the 'tank', 'healer' or 'nuker' roles. Perhaps they don't do it as well, but there should be at least a couple absolutely viable alternatives for any given 'required' role.

No Group Limits
By removing grouping limits, we can allow groups to fluxuate more naturally, and alleviate a lot of player-induced pressure to optimize the group. Due human nature, most players will not be able to create effective groups larger than 8 players. However, removing the hard cap on group limits removes the pressure to reserve open slots for 'better' groupmates or classes.

Super Groups
As described in the last post, most experience award distribution systems reward groups for being self-sufficient - further incentivizing them to min/max their members. This creation of an unwanted timesink, and social friction can be wholly avoided. In an informal solution, we could simply shift group rewards to be more inline with actions performed. If two groups cooperate to complete a goal, they should rewarded based on what they did - as if they were a single group.

Systems that focused on all-or-nothing rewards, and severe crowding cause 'kill stealing' and other anti-social behavior. If we split loot/experience rewards based on contribution, a group wouldn't necessary mind others helping them out. Ultima Online launched with a 'reward for work' system, and kill stealing was never an issue. Ninja looting was a separate problem though. It manifest because the UO system allowed players to be rewarded with loot independently of their contribution to victory. A loot system more like the one I've described previously would be more appropriate.

Formally, we could create a system which allowed groups to band together into supergroups. Indeed, it sounds as if Blizzard has caught on to this idea to ease the organization of raiding parties. With the creation of supergroups, we further reduce the stress on players to optimize group makeup, and are able to craft challenges and rewards more inline with the capabilities and contributions of each group. Imagine a higher-level zone like Dark Age's Lyonesse, with groups formed together taking on content together, rather than sectioning off spawns and 'camping' as individuals.

User-defined status monitoring
Allowing players to freely create a list of other characters to monitor (instead of only allowing current target and groupmates, as is the status quo) would also alleviate pressure from having to form the 'right' group. If a character can leverage their reactive capabilities (guarding, healing, buffing, etc) just as effectively on cohorts outside the group, as within, there is less 'need' to have someone of the required role in a particular group. So long as you have the right number of 'tanks' to protect the 'nukers', they needn't be in the same group.

Lax Group Level Limits
Anti-powerleveling code in most games needs to be toned down to lower organizational pressure. Again, players should be rewarded in the group based on the work they're actively doing. Only in rare cases should a player receive less experience for fighting a challenging target than they should get from fighting a level-appropriate target. Use-based systems (which reward players by increasing the abilities as they use them) demonstrate how much more casual grouping and sharing can be, while still avoiding the undesirable aspects of powerleveling. Level-based systems can achieve the same thing by granting experience to characters for using their abilities based on how often they use them, rather than how strong their opponent is. Avoiding a blindly scaled reward based on opponent difficulty helps in many more areas than just this one as well.

There are many other ways to try to address the problems of organizational downtime, and the pressures to arrange the 'right' groups. My suggestions are not for every game, nor every playstyle. What they are aimed to do, is merely make grouping a more fluid, more cooperative activity, performed via inclusion, rather than exclusion, in the interests of fostering teamwork and easing social friction.
 
  < / Organizational Downtime: The Suggestions >        
Thursday, March 25, 2004
  < Organizational Downtime: The Problem >        
Community being perhaps the primary draw of persistent worlds, we understandably take steps to encourage it, and make it visible. A large part of this, typically, is the requirement for players to get together into groups or raids, to take on 'fun' content. The unfortunate side effect of this, is organizational downtime.

Anyone who's been on a raid before knows (and loathes) the half-hour or more of listless waiting that occurs while players are gathered together, equipped, grouped, etc. Even in a regular grouping scenario, players spend quite a bit of time prepping, buffing, organizing, and waiting for other players. A large part of this downtime is strictly meta-gaming. The actual time required to heal/buff/equip, is miniscule in comparison to the time required to organize 'properly' in accordance with game mechanics.

Organizational downtime is the primary barrier to experiencing some of the best content persistent games have to offer. We would do well to understand and recognize when and where we create it, and what we can do alleviate it. First, let's take a look at what common design features are the most regularly appearing causes.

Required Classes
In most games, certain support classes are all but required for effective grouping: buffers, tanks, crowd control and healers primarily. A group is massively less effective when it isn't stocked with the right mix of core roles. A large selection of available 'hybrid' classes, existing outside these roles, expand the number of players who won't fit the core roles. These hybrids must be more carefully metagamed into appropriate groups to ensure proper coverage. Certain classes even do their best when grouped in bulk with perhaps a single healer -- rogues and scout-types for instance. When classes are so dependant on each others abilities, there is heavy emphasis on getting the 'right' group composition before adventuring. This takes quite a bit of time and regularly excludes players and even entire classes.

Group Limits
Relatively low numbers of players per group can exaggerate the problems of required classes. If there are 3 absolutely necessary roles to be filled (E.g. - CC, Healing, Tanking) then any given group in most games is already half spoken-for. If only a handful of available classes fulfill these required roles, we wind up with the majority of players vying for the minority of group slots. This inevitably leads to frustration with hybrid classes, and a general feeling that some classes are outright 'useless' (arguably true). A group limit that leaves a small number of slots open after the 'required' roles are filled will exacerbate the aforementioned metagaming downtime.

Experience Award Distribution
To counter the phenomena of Kill-Stealing and Loot-Stealing, most games emphasize the reward received by groups for damage done to a target when they are the sole damagers. When groups split effort on the same target, invariably both groups receive less experience combined than either one would have received, had they taken out the target all by themselves. In games with stiff competition for content, this manages to fulfill its purpose.

However, as soon as the game becomes cooperation for content - these systems cause problems. Groups are further incentivized to be self-sufficient, as several special-purpose-groups cooperating on targets advance much more slowly. This penalty is in addition to the further burden of requiring excellent players to compensate for missing 'roles'.

A social solution to this problem has been achieved in many games with cooperation for content. Groups within a raid are educated and socially incentivized to each stick to a single target when possible. This further adds to organizational downtime and bumps up the required education for new/casual players.

Group-Only Status Bars
Support classes generally enjoy being able to see the status of those in their group. Being able to see which buffs are fading out, already on, and whose health/mana/stamina is low allows them to leverage their abilities more effectively. However, when we limit this information to only being available for group-members, we emphasize the necessity to separate these classes into their own groups. Furthermore, we create educational roadblocks for newer players as multiple buffers/healers learn to split up their workload -- and this knowledge must be transferred. Not being able to formalize an interface so that healers can selectively choose who they can monitor (in or out of group) lessens the capability for 'odd' group compositions to be effective, on top of the other penalties.

Hard Group Level Limits
Anti-Powerleveling code also performs its purpose at the cost of increased organizational downtime. Groups must not only be class-balanced, but level-balanced as well. Depending on the existing group composition, an otherwise suitable addition might push other members of the group into the anti-PL experience penalty merely with their presence. Such requirements further emphasize required class downtime, and add their own layer to the metagame. Broader ranges of levels allowed without the PL experience penalty lowers the metagame timesink, but also lowers the effectiveness of its intended purpose: to prevent powerleveling.

There are certainly other systems which cause metagaming downtime. The above are merely the most common and egregious offenders. Next, I'll delve into my suggestions for lessening these effects.
 
  < / Organizational Downtime: The Problem >        
Tuesday, March 23, 2004
  < The Change-Up >        
Recently I've come to realize that my posts are rather long and digressive. A big part of this is that I wish to avoid predictable responses by covering all my bases before presenting an idea. The last thing I wanted to do was to have a suggestion in favor of level-free design elicit a few dozen accusations that 'skill based design has been around forever' and things of that sort. Yet this approach is making my updates much too long for everyone, and the 'preface' to my suggestions is getting painfully repetitive to write.

So to change things up, I'm going to aim for more concise, and hopefully more frequent posts. I figure if I'm not wasting time rewriting the same justifications over and over, I can put out more content with the same free-time investment by me. I'll also be making the post something clearer and easier to read, that isn't 6 pages, single-spaced. Any larger topics I'll likely break up into problem identification, suggested solution, and examples.

The latest (regrettably late) post (Dynamic Content vs Predictable Play) is roughly the length I'll aim for -- minus a few paragraphs as the plan goes. If anyone has any questions/comments/concerns on this, feel free to drop me a line. However, since most of you are either mute or fabrications of my tortured mind, I don't imagine I'll get much feedback either way.
 
  < / The Change-Up >        
Friday, March 19, 2004
  < Dynamic Content vs Predictable Play >        
The primary reason designs keep coming back to static camps, is to serve as predictable play. The primary player complaints with dynamic content designs (UO/SWG) is the inability to consistently find 'Fun'. The lack of a dependable known 'point of adventure' has players wasting much of their time wandering about, looking for fun.

Static camps serve the goals of predictable play, but they add a further removal of verisimilitude, a perception of scarcity, and monotony. The inevitable crowding around static camps has players wasting much of their time sitting around, waiting for fun. Trying to add 'more' content to alleviate the perception of scarcity is largely fruitless. Even when there are two or more zones which provide suitable content for a given character-power-range, players will congregate in only one of them. Early powergamers will determine a slight advantage that one area holds over the others, and a social standard will be born.

The critical mass of players in that 'better' area will also make it the better place to find groupmates, trade, recruit for guilds and socialize amongst peers. Once the community has 'set up shop', that zone will offer the lowest organizational downtime and the most feeling of community on top of its other benefits. Organizational downtime itself being one of the larger timesinks in persistent games, this makes the advantage absolute.

Mythic's 'camp bonus' was an honorable attempt to promote group movement. Basically they awarded extra xp to players for fighting at a spawn they had not fought at recently, and the bonus would slowly erode to 0 the longer they stayed. It was hoped players would more often rotate through the spawns and venture to under-utilized content. What it demonstrated was that organizational downtime incurred by moving between points of adventure was a larger factor than the camp bonus. Finding a new, unoccupied, camp spot, rallying the group to the new location, and determining new tactics provides more penalty than the camp bonus offsets.

Powergamers however, with their strict organization, were making good use of the existing camp bonus and were leveling quite fast in underutilized zones. Increasing the bonus only further helped those who already did well, but did little to nothing for the bulk of players.

Instancing, the new design rage, is thought to better serve the goals of predictable play. Interestingly, it provides essentially the same gameplay experience that text MUDs have had for decades. In MUDding, the concept of raiding a zone, leaving to regroup, waiting for the 'repop' and doing it all over again is nothing new. In most MUDs its the modus operandi for all PvE adventure. The question instancing raises is not one of mechanics, but of philosophy. Being able to consistently find content is desirable, but is it enough to grant easy access to such repeatable play?

Lubricating adventuring with instancing speeds up the level of activity, but will emphasize the lack of context we provide for our conflict and 'adventure'. Will making context-free adventuring easier be enough to draw in players? Or will it ultimately just make the experience more convenient for players who already put up with camping? Is there a better option?

SWG mission terminals make for interesting study. They provide players with a dependable launch point for the particular type of play they're seeking out. They can be balanced to spread players out amongst the terrain, and provide an appropriate reward that takes travel time into account. However, like instancing, they are merely a lubricant for adventure, and leave players with the hollow rewards of context-free (or at least context-light) conflict.

Helping a farmer defend his moisture vaporators does not net the player a newly friendly resource in the game world; it merely garners some XP and credits. Most times in SWG the quest-giving NPC is purely a story abstract, not even necessarily part of the game world itself. There's no farmer's political organization to curry favor with, and the nasties can never be successful in destroying a farm.

But why not mix the concept of mission terminals and truly dynamic content? Why not have NPCs or bulletin boards where players can be dependably directed toward desired types of content? There is a problem in this simple solution for any playtype that involves sending the player after consumable content (a monster that can be killed, a resource that can be depleted, etc). What if a player sets out to kill some goblins, and while they're en route, a second player happens upon and kills the goblins before the first ever sees them?

SWG's mission terminals alleviate this by simply not spawning the consumable 'target', until the player is appreciably close. This again removes the encounter from the context of the world state, yet a compromise is workable. Why not have the system log the position and type of the content that the player agreed to go find. When the player reaches the point where the content should have been, the system can ensure it's still present. If it had been eliminated or wandered too far off, while the player was traveling, the system could simply spawn more.

This way, targets wouldn't generally pop up completely outside the context of the world state, and when they did they'd be anchored by plausibility. After all, there were a bunch of goblins right there a little while ago right? Though these spawned goblins technically weren't, it's at least plausible to have a second group in the same area a short while later. Yet the location of adventure and the consumable target are still tied back to the world state. Should a group of players remove all goblins from an area, the mission terminals wouldn't be happily spawning band after band of cookie-cutter goblins for every other Tom, Dick, and Harry.

To further minimize even this possibility for a slight divergence from context, we could sort potential consumable targets according to how many players are in their vicinity. This way players will be even less likely to ever witness such reduced-context 'spawning'.

We could even get truly fancy with our mission terminal system, and require actual NPCs to 'witness' content (find a resource, spot some goblins) before the content would be listed on the terminals. We could even allow other PCs to 'report' content to the same quest-giving mechanism (though that would be trickier and require quite a few checks to avoid easy abuse of the spawning system).

The desire to provide players with a dependable way to find 'Fun' is a good one. Yet we needn't necessarily compromise a dynamic content environment to do so. Planetside, while its goals and requirements are quite different from traditional 'RPG' gameplay, has a quite similar system in place to help its players find 'Fun' in their ever-changing PvP landscape. Needless to say, it is a quite popular mechanism, and one that is wildly preferable to a simple reduction of the possible places content could be found.
 
  < / Dynamic Content vs Predictable Play >        
Tuesday, March 16, 2004
  < Designs that Shaft the Casual Gamer >        
Most of the design discussions I have online end up coming back to my feelings on this topic. As such, I figured it would be beneficial to create a condensed overview of my previous ramblings on the subject for posterity and reference. What follows are ideas of mentioned in my backlog of posts, collected and concise.

CG : a casual gamer. Casual gamers are defined here purely by the rate at which the gamer advances characters over real-time. It is used regardless of actual playstyle. Casual Gamers may be players with lots of online time who are heavy socializers or split their play up between many characters. As almost no persistent world has managed to keep CGs, there is no existing hueristic for how many offline CGs there are in sum population for any one CG online.

PG : a power gamer. Power gamers are likewise defined regardless of playstyle. PGs advance much faster than CGs over the same period of real-time. Every PG online typically represents 3 PGs in sum population. As they are online much more often than CGs, they are the foundation upon which persistent world communities form.

To visualize the community/population dynamic between CGs and PGs, consider "The Cheers Analogy":

PGs are essentially always there. They are comparatively few in number, but they almost all run into one another regularly. They are represented by the core Cast in the Cheers Analogy. They are Sam, Frasier, Norm, et al.

CGs on the otherhand are more numerous, but their stays are shorter, and more infrequent. They are represented by the extras in the analogy. As the core Cast is almost always present, the extras all know Sammy, Cliff, Diane, etc. The cast will certainly know some of them, but the extras themselves will rarely know each other. It is straightforward to see then that PGs establish the tone and social norms of the community (pun intended). They are indespensible social anchors and community cornerstones.

Currently, CGs are not paying to play commercial persistent worlds. Box sales vs Subscriber base indicates that most persistent world gamers are simply shopping for the one persistent game. That game is the one that engages them so much that they are willing to become a PG in that title, and can justify the fee. The vast majority who might be CGs in additional persistent worlds are driven away by the mechanics which heap penalties upon them. These penalties are outlined below.


Harsh Death Penalties
Harsh Death penalties designs are almost always implemented with exceptions. 'Rez' Spells or abilities that mitigate these harsh effects are unlocked through progress. CGs spend proportionally more of their sum play-time at lower levels, where these abilities are unavailable, so harsh death penalties wind up shafting the CG. They will spend a larger portion of their total online time doing corpse runs, spend more of their in-game money buying back stat points, and waste more of their in-game adventuring working off experience penalties. PGs will have 'unlocked' these abilities and will very very rarely suffer these penalties.


Required Camping
Requiring players to camp a static point waiting for quest material shafts the casual player. PGs will nearly always be present 'first', and the CG will spend proportionally more of their playtime waiting for a single spawn. PGs have more free time available, more frequently play 'off peak', and are less averse to 'waiting in line' as they have more time available.


Static Spawns
Games with static spawns shaft the CG. All the resources a CG can check with, in-game and out, will point them toward the content that the PGs have identified as 'best'. If spawns are static, a repeat of the above scenario occurs. Most times the CG will find all the 'best' spots occupied and will spend a disproportionally large part of their playtime simply trying to find a place to adventure.

Going to an alternative adventuring area is a penalty in and of itself. There are fewer others to group with, fewer others to trade with, fewer others to call to for help/heal/buff/rez, and an overall removal from the community. Using alternative areas makes advancement more difficult and excludes the CG from the community at large, and crowding further slows down the already tragic pace of advancement for the CG.


Bigger Carrot
Steep linear (if not outright exponential) scaling of loot over character advancement stratifies the economy. CGs fall into an economic stratus below PGs as they fight less and harvest less. CGs are thus unable to collect enough money to be viable trade participants with the PGs. They cannot afford to pay meaningful amounts for 'hand me down' gear and they do not get any gear that might be useful to sell up the chain. Level-based reward advancement punishes those who cannot keep up.


Crafting systems
Level-based design also results in a stratified crafting economy. Crafting systems are nearly all designed so that crafted goods are worthless up-the-chain. High-level goods require high-level components, and high-level adventurers are penalized for using low-level craftables/consumables. As the CGs are online more infrequently, CG crafters also have less online competition at any time (and fewer co-crafters to share cost/profit), so CG adventurers wind up paying more for crafted goods than the PG players did at the same level range.

Combined with the Bigger Carrot problem, this results in total economic stratification of the playerbase. CGs are entirely economically worthless to PGs, and as such are completely ignored by them. As the PGs form the cornerstone of the community, this alienates the CGs from most of the people they see online regularly, and would otherwise look up to.


The Required Monthly Fee
Requiring gamers to pay for a monthly fee, regardless of online time, alienates casual gamers. If a design has players bashing insects and vermin for 80 hours before they can engage in 'cool' content, PG's will reach that 'cool' content for a single monthly fee, whereas casual gamers will have to pay at least twice as much. Advancement doesn't have to be instantaneous, but a simple, rough, 'per hour' micropayment system as an option could correct this problem. (relevant suggestion)

One of the primary reasons players of commercial persistent worlds generally play so much, is that those who don't play an awful lot are much less willing to pay for the game. Similarly rare are PGs who play multiple persistent games. At the very least, casual-play friendly payment would increase the frequency of multiple accounts.


Scaling Time Sinks
Time Sinks for recovery, travel, and crafting tend to scale with advancement. That is, it takes longer for a level 10 player to heal naturally than a level 5 player, longer to craft a level 10 sword than a level 5 sword, and longer to travel from 'town' to a suitable adventure area at level 10 than level 5. To adjust for this, nearly all persistent games implement travel-aiding and recovery-aiding abilities such as health and mana regen, movement speed, etc. As these helpful abilities are generally reserved for higher level content, they do mitigate the effects of timesinks on powergamers, but leave casual gamers floundering in the low-levels, and suffering particularly in the levels just before these abilities are received. In general, higher level characters will have less required downtime between fights.

Crafting timesinks don't get adjusted for, but their steep linear increase keeps all but the most dedicated PGs from advancing. Steep timesinks also incentivize crafters to churn out junk no-one wants, rather than actual usable goods. This keeps prices for usable goods higher and clutters the world.


Limited Group Sizes
Hardcoded group size limits incentivize gamers to stock their group with the most effective participants possible. As the major drawback to grouping is organizational downtime, those who know the game, know their class, and are going to play longer sessions are the most desireable. This naturally leads to CGs being treated as second-class citizens as they doubtlessly don't know the mechanics as well, aren't as well 'tweaked', and are more frequently coming and going from the game world. Already in commercial games with a heavy emphasis on group advancement you will see PGs looking specifically for people who have a certain minimum number of hours to play. They have learned that adding a CG who's only playing for 45 minutes isn't worth the organizational downtime.


Anti-Power Leveling Code
It is felt that rapid advancement cheapens the achievement of advancement itself. If it formerly took players 40 hours to reach level X, then any mechanism which allowed a 'newbie' to reach level X in half that time will be seen as invalidating much of the investment of higher level players. Anti-powerleveling code generally achieves this goal, and results in stratifying the playerbase.

Low-level players are wholly worthless to higher level players. As PGs level much faster than CGs, the CGs quickly become, literally, second-rate citizens of the community. Even if a CG happened to have grouped with PGs, the difference in advancement rate will quickly obviate his social contacts as the PGs out-level him.

The resulting community of CGs suffers longer organizational downtimes as they are forced to group with one another and deal with each other's frequent comings-and-goings.


Single Axis Advancement
In single-axis advancement all rewards for the character advance simultaneously. The single axis of advancement identifies a small subset of activities which ultimately lead to every category of reward. Becoming better at this one activity (fighting monsters) results in being rewarded with seeing cool places, getting cool loot, having more money, doing more involved quests, fighting cooler/bigger/smarter monsters, etc all at once. As designs seek to lend 'meaning' to advancement (primarily through time investment), they weight advancement considering all rewards, making certain that those with less time see/do/gain/create nothing 'cool'. Multi-Axis advancement would allow CGs to focus their time investment so that they could achieve some 'cool' rewards in at least one category.



Note that not all of these items must be corrected to draw CGs into persistent worlds. A few (monthly fee, economic stratification, social stratification) must be fixed, but the rest operate as a cumulative punishment that keeps CGs away.
 
  < / Designs that Shaft the Casual Gamer >        
Thursday, March 11, 2004
  < Archetypes >        
The archetypes in persistent games are well established. Regardless of setting, or genre, there is common typecasting amongst character roles. Whether skill-based or class-based there are assumed balance considerations that are adhered to. The particular mechanics of a game can see a 'new' archetype emerge, but the 'balance wisdom' will still be applied.

Balance boils down to: character capability should be a zero sum game. No benefit should be without a corresponding penalty. This body of archetyping balance wisdom is summed up as: characters should have distinctive 'roles'. This is achieved by exaggerating their capability and accordingly their penalty to favor one aspect or another of the game.

Primary ranged damage dealers should have their melee defensive capabilities penalized. Stealth characters should be nearly unstoppable when fighting with surprise, but should be weak in all other situations. Specializing in one area (ranged damage) is done at the cost of being prohibited from specialization in other certain areas (healing).

Current archetyping systems play out more like rock/paper/scissors - where the player's decision of 'weaponry' stays with their character forever. One can either seek out those they are specialized to defeat (paper to rock) or engage in a battle of attrition with those they are evenly matched against (rock vs rock) or be hopelessly outmatched by those designed to defeat them (paper vs scissors).

Traveling without a group which covers all roles is an ill-advised decision. Archetyping games may have a larger domain of specialists than just rock, paper, and scissor, but the effect is the same. Not fighting with a 'main tank' is akin to marching only scissors and papers into a stronghold of opposing scissors (pardon the absurd visual). Most groups will become homogenized out of need, and many attempts at hybrids or diversity will fall by the wayside.

It may seem as if this common wisdom is necessary though. What game could be balanced if a single character could 'mez' a group of opponents and wield the power to do massive damage? What 'assassin' could be balanced if he could survive a head's-up, toe-to-toe fight with a specialist soldier? How much fun would rock paper scissors be, if we introduced 'bomb', which defeated all 3?

The problem in my mind is one of scale. If the wizard wasn't all powerful at range, he wouldn't need to be so weak in melee combat. If a sniper wasn't gifted 'one shot, one kill' when fighting on his terms, he wouldn't need to be so handicapped in a more fair fight. The exaggeration of 'roles' or archetypes is what drives the system toward 'Rock Paper Scissors' balance.

A true alternative system would look more like Chess. Not to offend Rock, Paper, Scissors enthusiasts, but the game is quite a bit more simplistic than Chess. Chess has the advantage of limiting the occurrence of 'powerful' roles. If the analogue was extended directly to a persistent game's class selection, every player would surely play as the 'Queen' piece, and diversity would be out the window.

Yet the concept can be applied in aggregate form. Consider the pieces of a chessboard as constituent capabilities of the character herself. She has a large group of pieces used primarily for defensive posturing (pawns), several specialized mobile groups (knights, rooks, bishops), and vulnerable but powerful trump cards (king, queen). Why can't a character be composed of a collection of capabilities, which can be balanced not only by power, but by occurrence?

The main limitation of Chess, boils down to accessibility. The rules and speed of the game are not very inviting to new users. If such complexity could be made easier to grasp at the outset, with a good bit of its complexity intact, we could have a winner.

It's fairly clear to see that this suggestion is starting to describe the underlying mechanics behind Collectible Card Game (CCG) systems. It's no surprise to me that CCG-based systems are rising in popularity and cropping up more-and-more-often in both tabletop and computer games. These systems create more complexity and strategy than our archetyping, while retaining a good bit of accessibility. The economics of CCGs are topic for another time -- their strength as a balanced system is the important aspect. It's also important to note that their conceptual presentation (often as simply a deck of cards) is not a requirement. The positive aspects of these systems can be applied to any other conceptual system.

We could applying a CCG or chess-styled concept to a persistent game through 'class' capabilities. 'Roles' could be preserved, but by player choice according to imminent need, not just once at character creation. Consider each character as having access to all capabilities as a chess player has access to all pieces. These abilities could be selected as needed into a sort of 'active deck', if you will. Abilities would be balanced by effectiveness, timer, and 'cost'. A single powerful ability may take up as much space in an 'active deck' as two or more weaker abilities, or it may have a longer reuse timer.

A party could still have archetypal roles. Indeed the best parties undoubtedly would. The difference is that they would just be defined through play, instead of at character creation. 'Useful' roles would be more fluid and grouping would be more natural without the restrictions of required roles. Some roles would be more rarely realized, but no player would be shafted day in and day out for having selected to play one.

Such systems are naturally not perfect for every game nor every gamer. There's also no reason a hybrid system somewhere between CCG and archetyping couldn't be made. They are simply an under-recognized alternative to the balance/personalization problems that archetyping systems fall into. To appeal to the broader range of gamers, the diversity allowed by variable deck CCGs, and the accessibility are a natural fit. 
  < / Archetypes >        
Tuesday, March 09, 2004
  < Rated T for ... whatever that means. >        
This post will take a slight detour from my usual focus on persistent world gaming. Instead, I'm going to spend some time on a broader hot topic -- acceptable content in video games. In short: I think it's all bullshit. The ESRB is almost as ridiculous as the reactionary groups and the politicians.

The broader issue is highlighted by Comic and Cartoon industries. Those industries in their youth decided to draw lines regarding what was 'acceptable' content in their mainstream product. The net result was that their mainstream product stagnated and was derisively dismissed by generations as 'just kid's stuff'. Aside from solid indie content in both industries, the contempt was deserved. By pretending that some things should just not appear hand drawn (sex, drugs, graphic violence, etc), all they had left was 'just kid's stuff'.

The game industry is now at a similar crossroads. The gaming public understands that mature game content is a viable form of entertainment, if not capital-A art. Others, however, are attempting to categorize video games in league with the Saturday morning cartoons. The film industry avoided such content typecasting by adopting its voluntary rating system. While imperfect, this scheme has allowed them to nurture their industry into a mature form of expression and entertainment. Where the individual products are allowed to stand on their own merits, and the industry itself takes responsibility only for making general warnings about potentially offensive content.

The ESRB on the other hand, is not doing its job. We've heard the arguments, and read about the lawsuits. We've seen the egregious offenders come and go, and the maelstrom that hovers over others. But why? Bay and Bruckheimer don't cause these kinds of stirs anymore even when their films showcase drugs, murder, guns, foul language, willful disregard for law and morality -- even huge dead fake tits. For all its faults the MPAA's scheme deflects such threatening uproars. Sure, occasional content gets protested or boycotted - but the arguments focus on the context, not the medium.

Now why does gaming even have its own content rating system? Why doesn't the ESRB have the same categories and procedures that work in the MPAA's existing scheme? Why fight the uphill battle of 'education', when the film industry already did all the hard work? Why not just adopt the MPAA scheme wholesale?

Are developers and publishers afraid of what might happen if 'R' or 'NC-17' is slathered across racks of their best-selling games? Let's be honest, even Medal of Honor's contextual violence would pull an R, to say nothing of the mayhem in most other games. Are retailers unwilling to properly obey content guidelines regardless of the system? Is there some other roadblock that prevents the game industry from correcting the self-inflicted gunshot wound to the foot that we call the ESRB ratings?

Naturally, the MPAA's rating system isn't a perfect fit all by itself. There is after all a difference between the media. Simply adding some of the content clarification we already find on cable and satellite TV would easily bridge the gap though. Why not label Vice City: R - excessive violence, mature themes, brief nudity? Is that somehow worse than 'M'? It's certainly more honest and more instantly accessible. Hell, at that point the ESRB could keep its silly system -- just adding the content descriptors would be enough.

No system could please everyone - but all this negative energy focused at the medium could be avoided. All that's necessary is to give the consumer an honest assessment of what's inside the box. There's no stopping groups from deciding to protest the line 'Kill All the Hatians' while happily ignoring the number of Cubans the protagonist is directed to kill. There's no stopping kneejerk litigation from trying repeatedly (thankfully, still unsuccessfully) to sue iD Software for a decade-old game. Senator Lieberman, however, could have his crusade directed toward those who ignore the industry's warnings (retailers, parents) and away from the medium as a whole.

The ESRB scheme, as is, doesn't even manage to properly label content. If ever a mainstream game deserved the AO rating, it was Manhunt. If the ESRB couldn't honestly rate that game, it only gives credence to these moral crusaders. Looking at the M rating on Manhunt, it's hard not to feel like maybe the industry isn't making an honest effort. That kind of track record only makes things worse.
 
  < / Rated T for ... whatever that means. >        
Monday, March 08, 2004
  < AFK >        
No post this past Thursday. Took some AFK time. 
  < / AFK >        
Tuesday, March 02, 2004
  < A Run of Bad Luck >        
'Chance' is a big theme in the modern roleplaying zeitgeist. Nearly every game system that models conflict includes a random constituent in all its resolution checks. Characters have a 'chance' to hit, dodge, sneak, climb and accumulate word-on-the-street. Their abilities and the situation merely modify this chance. It's a concept that was created and embraced as a 'good enough' stand-in for complexity. Rather than modeling the exact positioning of two opponents, we say there is a 'chance' of hitting a particular location. We embrace it on the tabletop, as it replaces mathematical complexity with a mechanic that engages the player and keeps the game rolling (pun intended).

As with many of the holdovers from Pen-and-Paper, I think this particular concept has largely outlived its usefulness in our computerized niche. The effects and merits of various random number generation algorithms have long been the subject of gamer and designer study. The one thing that nearly every student of randomization takes away from this study is: players do not expect, nor appreciate true randomness. When a player sees a 5% chance of critical success, and a 5% chance of catastrophic failure - they don't expect such events to follow each other 'back to back'. Such 'streaks' are a necessary requirement of a true random distribution. Yet it seems players expect a normalized distribution (one with a familiar bell curve). As a result, most modern games have had their random number generation systems tweaked to follow a normalized distribution, or had their 'random' components drastically minimized in importance.

Beyond that is the problem of the common gamer perception of chance in general. Good fortune on 'a roll' is commonly internalized to the gamer himself: "*George* sure was lucky that round", "Damn *I* got lucky there". At the same time, misfortune is regularly externalized: "*The game* is screwing me.", "*These dice* are bullshit.". We can see generally, that players perceive bad luck to be a fault of the system. When they have good luck, their positive feelings are attributed to themselves, not the design.

So why should we use random numbers at all? If we choose to use it to determine 'chance', then why don't we bring good luck down from the metagame and add it as a direct game mechanic? Perhaps 'luck' is an ability that allows a character to modify one of his abilities temporarily? This would cause the playerbase to correctly associate 'good luck', or the lack thereof, on the character; not the player nor the game itself. Surely in doing this it would be important to ensure that not everyone has luck, and that luck isn't too effective. If we erred in this design the very lack of 'good luck' would be synonymous with permanent 'bad luck' - which is wholly undesirable.

What about the use of random numbers to simplify the complex parameters involved in resolving conflicts? As we're using computers, we can model all sorts of complexity with no adverse effect. Often, we already are modeling enough complexity that dropping randomness wouldn't cause much of a noticeable change in gameplay. DAoC exemplifies this in the 'DPS' ratings they give to their the weapons in-game. Mythic is outright advertising that, despite some variation, the player will experience a normalized distribution around an average 'damage per second' output.

So why do they bother with any randomness at all at that point? Granted they use character abilities to modify this distribution -- higher stats mean a tighter distribution toward the top-end of a weapon's potential. Still, the effect is that they have an output average that is guaranteed over even small quanta of time. So what's the benefit to retaining a computationally expensive random component, or the even more expensive distribution-correction code?

The common response is that without 'chance' - combat is stale. The player will always know how much damage he will do and take, and conflict resolution becomes more an exercise in math than tension. Think for a moment though on your last massmog: players are already playing the numbers game. A player quickly learns what kind of damage they can count on dealing and taking from a given mob. This 'chance' we crave is carefully minimized during development to ensure that it is one of the more minor factors involved in adjudicating conflict. But by having my character hit for 5, 9, 6, then 8 -- am I really gaining something over simply hitting for 7 each time? Indeed by removing this chance altogether we could more clearly demonstrate the nuance and strategy in the complex systems we already use.

Indeed the way to make combat fun and engaging is in varying the opponent, application of tactics and situation. Is Risk inherently more fun than Chess because of dice? Granted they are very different games, yet it doesn't appear that the inclusion or exclusion of dice significantly impacts the enjoyment of either.

Instead of spending time and effort creating a false approximation of chance, why don't we code systems that re-emphasize tactics and strategy? Indeed most of the systems we implement now, special moves, timers, counters, etc - are already doing vastly more to keep combat flavorful than random numbers ever did. Players are more engaged by being able to form strategies than by sitting with their fingers crossed.

This is not to decry all randomization. Indeed there are times when 'random' is beautiful and constructive; When the requirements and properties of true randomization are all positive. For instance: cloud formation, variation in NPC appearance, terrain details, etc. These are systems in which the effects of streakiness and true random variation is desirable, while normalized output is bland.

Going back to player perception of chance - we can see that in Player vs Player conflicts, chance is going to always be met with derision. One player out of a two player conflict decided by chance will blame the system for their loss. If a particular fight isn't ultimately decided by our random numbers, then their inclusion was pointless. By removing this 'chance', we re-emphasize the importance of tactics, timing, and ability. More importantly, we reinforce that combat is determined by ability and strategy, and not dumb luck.

Instead of systems that grant a percentage chance of total success or total failure, why not use systems that grant partial successes? Consider D&D's 'save for half damage' mechanic. Each player has a chance to take a lesser amount of damage from a spell, poison or trap. Chance dictates whether the character takes half damage or full damage - a statistically significant amount to take out of the hands of the player. Wouldn't the player be better served if we turned a 50% chance of half-damage into a certain 75% damage? Why not graduate success/failure instead of leaving so much up to a pass/fail chance?

This will remove the focus from luck, and put strategy back to the forefront in conflict. We don't (and shouldn't) do this on the tabletop, because determining fractions of damage for every attack or resistible spell would be cumbersome. Yet with computers doing the dirty work it is actually faster to scale damage according to ability than to determine a random number, let alone correct it for a normalized distribution.

I say its time to toss out random components from our conflict resolution mechanics. They're slowing gaming down instead of speeding it up. They're pulling players out of the game rather than keeping them engaged. They aren't a solution anymore; they've become a problem.
 
  < / A Run of Bad Luck >        

Powered by Blogger