Role Playing Game Design

   For a look at role playing game design, we’re going to break down an existing system – in this case, Legend of the Five Rings (Third Edition) – and take a look at where it seems to build on prior systems, where it derives from basic theory, and where it innovates.

   Like the vast majority of role-playing games, Legend of the Five Rings has attributes (measuring a character’s innate potentials) and skills (measuring the things he or she has learned), weapons and armor, an initiative and damage system, an experience system (allowing the characters to improve over time), a system for determining whether or not the characters succeed at various tasks, and a magic system (some games call this a psionics system, or “technology”, or a dozen other things – but they’re all rules for letting the characters do things that are physically impossible). Unlike all but a minority of games, Legend of the Five Rings (or L5R for short) includes formal systems for social effects.

   We’ll take a look at all these elements to see where those game elements came from and how they were developed.

Attributes and Skills:

   To start with the basics, the attributes are a fairly standard set for a role-playing game – four physical attributes, representing muscular power, endurance, manual dexterity, and agility/reaction speed, paired with four mental attributes, representing alertness, social perception, intellectual ability, and force of will. They’re paired up to link them to the four classical elements – Fire, Earth, Air, and Water – in a fashion rather reminiscent of the subattributes from the Players Option series (1995). To fit the oriental theme, and the classical Book of Five Rings (circa 1645) and it’s classification of all forms of combat into the books of Earth, Water, Fire, Wind, and Nothing, there’s a fifth element thrown in – Void – but it’s really an attribute value all by itself, since it has no subattributes. It could be taken to represent luck, chi, or spiritual strength with equal validity.

   The values of the elemental superattributes are set by the lower of the two attributes that make them up, unlike – say – Big Eyes, Small Mouth, but that does help keep the characters development a little more balanced than it might be otherwise. Overall, the physical/mental division, and even the pairing, is flavorful, but not new. It also makes Void the cheapest of the five superattributes to raise, but living in a world where magic actually works might justify that.

   The skills are also a fairly normal set, divided into Courtly, Warrior, Mercantile, and Peasant skills, reminiscent of Bushido’s (1981) division of skills into Fine Arts, Combat, Practical, and Ninja skills, or of first edition AD&D Oriental Adventures (1985) division of proficiencies into Court, Weapon, Artisan, and Common groups. The specific skill names vary, but the functions are all recognizable from many other systems.

   The L5R system also gives each of the skills both generic – for every skill – and specific bonuses and enhancements to the user’s abilities at various levels, a fairly classical idea I first ran across in The Complete Adventurer (1983). Unfortunately, while this does make each skill unique, it means that the characters gain a cookie-cutter set of abilities from them. It’s also extremely inelegant – demanding a unique write-up for each skill. That’s occasionally justified for supernaturally-based skills and those governing odd abilities, but it seems a bit excessive for skills like “acting”.

   Like the original Vampire the Masquerade (1991), both attributes and skills may eventually be increased to a godlike 10, although starting characters will normally have far lesser totals. In fact, the system uses the same basic scale as Vampire: The Masquerade: 1-5 for normal humans with 2 being average and 5 being exceptional. The rest of the numbers up to 10 are available for great heroes and supernatural creatures. Why have “normal” towards the lower end of the scale? Simply because – like most games – the system is more interested in heroes than in ordinary folks. That’s fine of course, so are most players, game masters, and readers.

   Overall, both the attributes and the skills are fairly basic sets, with too many possible sources of inspiration to trace.


The Experience System:

   Experience systems come in three basic varieties. Table-based systems track experience – whether as a generic item or in specific categories – and award bonuses at particular thresholds. They’re often referred to as “level based” systems, such as Empire of the Petal Throne (1975). They’re realtively easy to use and make it easier to keep the characters relatively well balanced, but are unrealistic. Usage-based systems gradually increment the skills and abilities you use, as in Runequest (1978). They’re somewhat complex, make it difficult to compare characters, tend to lead to everyone looking a lot alike in the long run, and make it very difficult to study new or unusual things. Point-based systems such as Champions (1981) hand out abstract points, and let the characters spend them buying the abilities they want. That’s relatively easy, but lets people build terribly unrealistic specialists or otherwise unbalanced characters – and means that what the character learns often has no relationship whatsoever to what they were doing. That’s one reason why hybrid usage-based/point buy systems, such as was used in World Tree (2000) became popular.

   L5R, uses the point-based system: you get a relatively small number of points per session, which you may spend to buy up skills, attributes, and – indirectly – to increase your “level”, and thus gain special abilities. Unlike Champions, which uses a linear cost structure to allow superheros, L5R uses the same progressive-cost system as Shadowrun (1989), TORG (1990), and Vampire: The Masquerade. You must buy each level of an attribute or ability in order, at a cost of (the level being purchased times a multiplier) – in this case, 1 for skills and 4 for attributes (TORG used 1 and 3, Vampire 2 and 4, first and second edition Shadowrun used a variable cost of 1-2 for skills – depending on how specialized they were – and 1 for attributes, as they were noticeably less important in that system, Third edition used .5 to 2.5 for skills, depending on how high and specialized they were and 2 for attributes).

   Overall, the only unusual item here is the systems computation of level or “Rank”: it’s an over-complicated mess, involving rings, skills, special bonuses from skills, and possibly special bonuses from natural advantages or other sources, thus making a (somewhat chaotic) attempt to reflect the ideal of a well-rounded individual. Each such level acquired allows the character to learn one technique – more or less powerful secret abilities resembling the martial and magical skills of Empire of the Petal throne. Characters normally pursue particular schools, a classical part of Japanese history which apparently first turned up in Land of the Rising Sun (1980).

   Another place where the system could definitely use a little work. There’s nothing wrong with limiting special abilities through one form or another of “levels”, but the implementation here is overcomplicated.


The Task Resolution System:

   Basic game systems tend to address only whether a given action succeeds or fails. Even some of the earliest games, however, began to introduce ways to go a little deeper than that. For example, Dungeons and Dragons, from very early on, had you roll another die for damage – checking for how well you’d done with your attack after first checking to see if it succeeded. Runequest added more levels of success – critical successes and fumbles. Non-weapon proficiencies, with a provision for inferior, average, and superior results were introduced to AD&D in first-edition Oriental Adventures, although the basic mechanic – rolling under the relevant score – was introduced in Bushido  years earlier, along with a general system of situational modifiers, critical successes and failures, and an extended action or “Task” system.

   Of course, the basic job of a game master has always been to adjudicate how player’s actions work out. That’s why – just as soon as dice wandered over into the original RPG of “lets pretend” (probably first played by pre-humans) – you started to hear phrases of the general form “Sure, you can try that – but at -10!”. The original Marvel Super Heroes game (1986) formalized this kind of adjustment by shifting columns, or even the required color of the result needed, on their universal chart. Dice pool systems that counted each die individually might shift the target numbers (like Shadowrun and early White Wolf systems) or call for extra successes if they used fixed target numbers (like later-edition White Wolf systems).

   Single-die and dice-total systems normally just shift the target numbers, scaling them to fit in with the expected results of the die system in question. In single-die systems the step size for such adjustments tends to hang around the 5-10% level – +1 on a d8, d10, or d12, +1-2 on a d20. In die-total systems – regardless of whether or not you’re totaling the entire pool – such adjustments tend to be about the value of one die, usually rounded down. That’s +/- 3 for d6 systems, 4 for d8 systems, 5 for d10, 6 for d12, and 10 for d20 systems. That’s an easy and quick way of scaling difficulties to the characters for the game master: each (+1/2 the die type) added to the difficulty equates to asking the character to have another die in the total.

   Getting target numbers – whether for totaling systems or for individual-die systems – is fairly simple. When you come right down to it, you rate some common tasks for difficulty and how often a character with a given set of attributes and skills should succeed, and either do some calculating or roll the dice a bit. Then assign target numbers to get the results you want. If a character is opposing another character, just let them roll against each other. That will set the difficulties for you automatically. Empiricism does have its place after all.

   If a character wants to try something fancy, adjust the target number up or down an appropriate number of steps, and let them go ahead and try. Some systems include an extensive list of how many steps should be allowed, or limit how many difficulty-boosts a given character can apply, others leave things entirely up to the game master.

   L5R falls in the middle: there are a few specified effects you can apply (mostly in combat, but some for creating items or modifying spells) at the expected half-a-die-per-step, but most modifiers are up to the game master.

   Like White Wolf, and later-edition Shadowrun, Legend of the Five Rings uses a dice pool system, with the pool derived from adding an Attribute to a Skill. The two major variants on dice pool systems are Totaling the Dice Pool, as was done in Tunnels and Trolls (1975), and Reading the Dice Individually – usually in conjunction with a Bonus Die/Exploding Die/Reroll Die/Wild Die rule, wherein each die that rolls the maximum is rerolled for an open-ended total on that particular die – such as in Shadowrun. Both are straightforward, used in many different game systems, and are relatively easy to analyze.

   Mildly unusually, Legend of the Five Rings totals only a part of the dice pool, allowing the individual rolling to select the retained dice. That’s nothing new – that was how “Method 1” attribute generation worked back in the First Edition AD&D Dungeon Master’s Guide (1979) – but it is unusual to use it as the basis for a system. Given that it is unusual, we’ll have to see how it derives from general theory.

   How to decide how many dice to select is the first question. Obviously, it should normally be less than the number rolled, yet greater than zero. A fixed number is meaningless. It simply goes back to using the grand total at most or all levels of (Attribute+Skill) if it’s too high. If it’s too low, high attributes and skills become less important. If you’re only keeping a few dice, it’s easy for a lucky throw on a few to exceed the results from a much larger pool – and you drastically restrict the potential for heroic feats. (Like it or not, heroic feats are one of the major attractions of RPG’s). So: That leaves us with adding a third variable (call it “aptitude”) which determines the number to keep, using some function of the attribute and skill numbers – usually the average (as in; “select half of them”, since the base mechanics shouldn’t be too complicated), or using either the Attribute Number or the Skill Number as the basis.

   Throwing in a third variable complicates things excessively (say, 10 possible aptitudes, 10 attributes, 30 skills… 3000 possible types of checks. A bit much) unless make it very general – and if we make it very general, it will affect far too many things. That’s why the god Thor, in 3.5 d20, is also a superhuman master of needlepoint. He has a set of bonuses that apply to all checks, not just to a limited set. That doesn’t work very well does it?

   Using the average is easy enough, but tends to eliminate the distinction between attributes and skills: if they’re both treated the same way. It’s also an extra math step, and forces you to come up with a rule on rounding – meaning that some normally-distinct values are effectively identical. Far more importantly, it means that someone with Attribute 1 and Skill 9 is just as good at all relevant tasks as someone with the opposite arrangement. That doesn’t work at all if it happens to be a roll about power lifting: skill helps with that, but not nearly as much as raw strength. In fact, that’s the general rule for a lot of things. Skill helps, but you need to have the basic capability to do whatever it is. Small children – lacking skills, but with some basic abilities – do better at physical sports than professional sports stars who happen to have suffered paralyzing injuries. Ergo, attributes are more important than skills – at least in (Attribute + Skill) dice pool systems. That’s a notable weakness for White-Wolf style games, but – luckily for them – they aren’t the topic today.

   The math shows that the number of dice you get to pick (Y) are more important than the number of dice you get to roll (X) as long as (X) is significantly greater than (Y). (The distinction no longer matters when (Y) = (X)). Since attributes are more important, and we’ve already got a dice pool of (attribute+skill), the simplest way to represent this is to let the attribute govern (Y). So: Roll (attribute+skill) dice, and select (attribute) dice with which to generate your total.

   Dice pools can become annoyingly large however. That takes us to Dice Pool Caps – a notion which has been around since the early days of Champions. When a player with a super-strong character doing haymakers needs to roll 24d6 for each punch, the cry of “take the average!” pops up very quickly indeed. Of course, a certain amount of variation was still desirable – which led quickly led to a “drop the tens” rule, wherein increments of 10 dice counted as being average and you rolled the remaining dice. Thus 24d6 became 4d6+70 Stun Damage (since this isn’t about Champions either, we won’t get into Stun and Body Damage).

   In systems where the dice are considered individually, matters are even easier: simply take the chance of getting the number you need – always less than unity if you need dice at all – and multiply it by the number of dice you’re discarding.

   Most systems with formal rules for capping dice pools round down, or otherwise penalize taking the average a bit: that’s because characters who automatically succeed all the time are boring – so the designers like to encourage rolling.

   Dice pool caps usually turn up at about eight to twelve dice – most often at ten. Why? Ten in particular because we use a decimal system for counting, making it a number that’s familiar, intuitive, and easy to multiply by – but the general range is simply because when you get past eight to twelve dice, it starts becoming awkward to hold, roll, and count them.

   The calculations on how this works are a bit harder when you introduce exploding dice or are being selective about which dice you count – but they’re doable, and like most things involving dice and calculations, you can find an assortment of tables and formulas in books on mathematics or on the internet.

   As it turns out, you can get a fair approximation of the effects of adjustments to (X) and (Y) in systems where you roll (X) and select (Y) to generate a total with with a few simple rules – if you’re always looking for a high number. Incrementing (X) is less effective than incrementing (Y), since you will be selectively taking the dice with the highest results anyway. Of course, (Y) cannot increase beyond (X) unless we go back to assuming an average or slightly discounted result for those dice we’re not actually rolling.

   What are the easiest set of approximations to adjustments in X and Y for relatively small numbers of dice? Well, a quick look at the various tables shows that:

  • +2X roughly equates to +1Y since some dice will usually be above average while others are below it – and you were already selectively keeping the higher ones.
  • +1Y roughly equates to +(the average die result). The closer (Y) comes to (X), the closer you get to the average, while the more the greater the ratio of (X) to (Y) the closer the effective value comes to the maximum value.

   Getting the average for “exploding” dice requires solving an infinite power series – but, for larger dice, the result actually isn’t much different from the base. We’ll use d10’s for this, since that gives us a decimal power series, which is really easy.

   Average of an Exploding d10 = 5.5 + 5.5/10 + 5.5/100, and so on. Add a few of those, and you’ll come up with 6.1111… Not really all that different from 5.5, although it allows the occasionally freakishly high result. The same sort of calculation applies to dice of all sizes though.

   Now quite a few RPG’s have foundered on requiring over-complex math – so we’ll stick with the simplest approximation and go with the rounding-down idea.

   Ergo, roll (X) and select (Y) systems you can readily limit dice pools to 8-12 dice. Each +2 dice you’d normally get in your pool over the limit equates to getting to select one extra die. Each die you’d normally get to pick over the limit equates to adding one-half the number of faces on your die type to the total.

   In the case of L5R, the easiest – 10 die – cap was used.

   Now, if you’re looking for a low number, or for precise control of the result, this won’t work properly. For that you want to have (X) as large as possible, while keeping (Y) small (for low numbers) or at an intermediate value (for precise control). That’s complicated to fix – the mathematics for precise control gets especially hairy – so it’s probably best to stick with the old “a higher score is always better” rule in your base mechanics. That rule is so old that it’s origin is lost in prehistory. Whether it’s an archery tournament or Monopoly, you want the high score.

   A bigger problem with basic resolution systems is that their skill results are usually all-or-nothing. That works fairly well if you’re up against an active opponent: if you tried a fancy trick with your sword and blew it, it’s not entirely unreasonable to say that you failed to effectively connect at all. If you failed to out-debate your opponent, you probably didn’t persuade him – and may well not have made an impression on anyone else unless they had a strong reason to agree with you already.

   It doesn’t work well for most unopposed tasks. In the real world, we have graduated levels of success. Simply because one failed to produce a piece of legendary craftsmanship that will serve as an example to lesser workmen for generations to come (despite your attempt to do so) doesn’t mean that you failed to produce a well-made item. Failing to figure out the precise composition of an alloy doesn’t mean that you failed to figure out that it contains a lot of iron. If you search a room, failing to find the metal shavings shoved into half an old sandwich in the trash doesn’t mean that you failed to find the handgun under the couch cushions.

   Unfortunately, L5R uses a simple succeed/fail mechanic for both opposed and unopposed tasks. I think that needs a patch rather badly, at least for our local games. Ergo, Patch One: If it’s a noncombat and unopposed check, you roll first, and get whatever results you can qualify for – although it’s worth noting, that even passive resistance from another individual means that it’s an opposed check.

   Now, player characters usually operate in a group – and when it comes to rolling for a group success or failure, there are only a few basic possibilities:

  1. Each character rolls independently and the worst result governs the outcome for everyone. For example, if you’re trying to sneak past a guard, the entire group is in trouble if a single character is spotted. Rolls like this are appropriate whenever a task has no margin for error – whether that’s performing a mystic ritual or trying to simultaneously defuse a dozen different bombs.
  2. An average roll is often used in the interests of speed or when it is assumed that good performance by some individuals can make up for poor performances by others. For example, in climbing a mountain when roped together, the experienced climbers can often assist the less experienced. If that’s a little too tense, the same sort of rule applies to carpenters building a house. There are an awful lot of situations where a team of experts will do better work faster, but where a mixture of experts and less competent assistants will still get things done. As far as the game system goes, about the only real questions on “average” rolls are “do we roll once using the average scores of the group or have everyone roll and then average the results?” and “do we round up or down” and- both of which give similar results. In most cases the practical answer is to (1) do whichever is easiest, and (2) to round down. After all, “almost” is not equal to “success”.
  3. Assisted rolls are used when the less-competent individuals in a group can effectively assist the experts – usually by working on limited or easier aspects of the task or simply by standing by to lend a hand when it’s needed. This covers cases such as an unskilled group assisting an archeologist: they can help lift, dig, and carry – and as long as they call in the expert when something comes up, all they need is to be able to recognize when they’ve hit something important. That lets them make a massive contribution to the overall project with fairly minimal skills, as students, enthusiasts, and local employees have done for countless archeologists.
  4. Having all the characters roll independently and using the best result. This works nicely when one character can succeed for everyone. For example, if everyone is trying to decipher an ancient scroll, individual failures have no effect as long as someone manages to figure it out (and is willing to share the information). Unless some serious supernatural effects are in play, however, it’s a bit much to expect the one fellow who is good at sneaking about to successfully carry everyone else past a guard.

   In games, cases (1) and (4) are normally covered by the basic rules. Case (2) is adequately covered by taking the arithmetic mean and rounding down. The complications show up around Case (3). Questions here include how many assistants can meaningfully contribute to any single roll, how big a contribution they should make, and how difficult the assistants check should be.

   L5R falls down badly here. While it does say how much any given assistant can help out – the usual one-half the die type for the dice pool (in this case a +5 since the system uses ten-sided dice like the various White Wolf games) – but it says nothing about how many people can help out on any given task and the mechanism for setting the difficulty of the assistants checks – one level / die less difficult than the difficulty that the leader has decided on – is decidedly shaky. If Sherlock Holmes could turn to unskilled street kids to assist him in his cases (an admittedly literary rather than actual example), why can’t the player characters? Did finding the knife in the shadows under the bed become more difficult because the group leader has decided to try and decipher the fragmented bits of code left behind on a crumpled piece of half-burned parchment? Will all the lesser clues collected by assistants simply vanish because the leader failed to find the strand of colored thread caught on a splinter in the framework of a hidden door?

   Obviously not. Ergo, we’ll want to patch this in our local games by falling back on the game master and the unopposed checks rule we just introduced. Patch Two: In an assisted skill check, the Game Master sets the target numbers for the assistants checks. No more than – say – (the leader’s Intelligence, or similar attribute) aides can effectively contribute.

   This article is becoming far too long for a single post: Ergo, it will be continued over the next few days.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: