Dice Pool vs. D20 vs. D100 (Percentiles)

[werescythe]

So, I've been experimenting around with the idea of trying to create my own game and doing my own rpg. While I was experimenting around with D&D 5e mixed with a few other game systems, I've kind of been reacquainted with percentiles as well as dice pool mechanics.

So, I wanted to ask a couple questions in regard to dice pool mechanics, the d20 system and systems that focus on percentiles.


What are the benefits of these systems?
What are the limitations of these systems?


Thank you all for reading this post. I hope you all have a great day.

[Anonymouswizard]

The benefits of dice pools is that they tend to produce predictable results. The more random elements you have the more likely you are to end up close to the mean.

Percentile systems' mainly benefit from your score being either equal to your success chance or 1% off (for those that treat 00 as 0 instead of 100). They also have an advantage shared by other roll under systems in that they can use a 'roll under, roll high' system, which is somewhat more cheat resistant.

Single die+modifier is easier to grasp. People find 'high is good' intuitive and it has an easy to understand probability spread.

[werescythe]

Okay, so then what are the limits when it comes to the stat restrictions? I know you can have negative stats in DND, like have an 8 (-1) in STR, but are you able to apply that with either the percentile system or dice pools?

What if you have a -1 for your dice pool?

[Yakk]

Percentile systems have the problem that people aren't good at probability.

Design wise, they have people wanting to use the range of values from 1 to 100. This (a) has insane granularity, and (b) is a huge range.

If you have played various D&D games, if human strength goes from 1 to 100, it would be like if humans strength in D&D went from -10 to +10 modifier (on d20 rolls).

The insane granularity means that you are communicating and recording data that is mostly noise. 17 vs 18 percent, on a roll or a skill, is almost always noise.

In almost any percent system I've used, unless you hyper-specialize you are incompetent in your "areas of expertise". A soldier with 30% gun skill in a system with penalties for range etc ends up with "can't hit the broad side of a barn". This continues the "low information" problem above; your turn is "I try to do X", and the answer is usually "you fail and nothing happens".

This isn't integral to a % system, but like I said, people are bad at math and probability.

...

Dice pool systems run into the same "people are bad at probability" problem and "lots of information that is discarded" issues. You'll get some kind of non-linear response in the system, and designers end up reverse-justifying why this non-linear response is desired. I haven't seen a dice pool system where the non-linear response is designed rather than justified; maybe the ORE engine by Greg Stolze. (It also attempts to deal with the wasted information problem by playing Yahtzee with the die values). Even then there are some annoying quirks.

...

d20's evolution from OD&D up to 5e has been interesting, because a lot of the above problems have been ironed out. Both 4e and 5e modifiers and the like where designed intentionally, unlike almost every other RPG engine I've taken apart.

[Maat Mons]

Ordinarily, in a dice pool system, the final number from adding all your stuff together is the number of dice you roll. You can't roll a negative number of dice, so if that's what the final math shows, it's normally treated as automatic success or automatic failure, depending on whether it's a system where more dice is better or a system where fewer dice is better.

In a system where you roll n dice and need at least one of them to come up a success to succeed, rolling zero or fewer dice is usually an automatic failure. In a system where you roll n dice and need all of them to come up successes to succeed, rolling zero or fewer dice is usually an automatic success.

Sometimes I like to contemplate a hybrid system. If your modifier is zero, you roll one die, and the success or failure of that die determines the results of your roll. If your modifier is positive, you roll modifier + 1 dice, and succeed if any of them come up as successes. If your modifier is negative, you roll abs(modifier) + 1 dice, and only succeed if all of them come up as successes.

But in normal dice pool systems, of the more is better variety, negative values just mean you need a bonus from somewhere else to even be able to attempt the roll. So if your skill is -1, you need +2 in favorable circumstances to have any shot.

It's worth noting that dice pool systems of the more-is-more variety typically never allow for a 100% chance of success, no matter how good you are or how simple the task. And dice pool systems of the less-is-more variety typically never allow for a 0% chance of success, no matter how bad you are or how difficult the task. The hybrid I outline above never goes to 0% or 100%.

It's also worth noting that, in more-is-more systems, rolls for highly skilled people are more consistent than rolls for low-skill people. But in less-is-more systems, it's the other way around. rolls for low-skill people are more consistent than rolls for highly skilled people.

[Anonymouswizard]

Okay, so then what are the limits when it comes to the stat restrictions? I know you can have negative stats in DND, like have an 8 (-1) in STR, but are you able to apply that with either the percentile system or dice pools?

What if you have a -1 for your dice pool?

In a dice pool <1 dice generally means either:
-automatic failure, or
-roll one (1) special die, generally with a minimal chance of success.

Both Chronicles of Darkness and the latest version of Paranoia use the second type.

In percentile and other roll under systems you generally won't have negative stats. Some value will be designated as the 'baseline' (I've seen everything from 10 to 60, but around 50 is I believe most common) and weak stats are below it with 0 as complete incompetence in the area. So you'll effectively have the 8, but there won't be a -1 modifier attached to it (generally).


Percentile systems have the problem that people aren't good at probability.

Design wise, they have people wanting to use the range of values from 1 to 100. This (a) has insane granularity, and (b) is a huge range.

If you have played various D&D games, if human strength goes from 1 to 100, it would be like if humans strength in D&D went from -10 to +10 modifier (on d20 rolls).

The insane granularity means that you are communicating and recording data that is mostly noise. 17 vs 18 percent, on a roll or a skill, is almost always noise.

Yeah, even as a fan of percentile systems I'm not really sure that a difference of less than 3 really feels like anything. The actual advantages are the ability for minor cheat protection, doubles being mostly evenly spaced (completely if you do 00-99 instead of 01-100), the ability to flip-flop rolls, and the ability to use the numbers on the dice or their total in addition to the percentile value.

The large range is mostly something you put up with.


In almost any percent system I've used, unless you hyper-specialize you are incompetent in your "areas of expertise". A soldier with 30% gun skill in a system with penalties for range etc ends up with "can't hit the broad side of a barn". This continues the "low information" problem above; your turn is "I try to do X", and the answer is usually "you fail and nothing happens".

This isn't integral to a % system, but like I said, people are bad at math and probability.

Game designers will game design, badly a large amount of the time. Just like how D&D in theory tries to make the average untrained total +0 many percentile systems could benefit from making the average untrained score 30-40 instead of 5-20.

I'm looking at you Eclipse Phase!

[Waddacku]

Yeah, even as a fan of percentile systems I'm not really sure that a difference of less than 3 really feels like anything. The actual advantages are the ability for minor cheat protection, doubles being mostly evenly spaced (completely if you do 00-99 instead of 01-100), the ability to flip-flop rolls, and the ability to use the numbers on the dice or their total in addition to the percentile value.

The large range is mostly something you put up with.
Many percentile systems end up working in increments of 5 or so, which leads to them effectively having the same spread as a d20 roll, albeit with immediately obvious probability of success and the mentioned tricks you can do with the digits. Regardless of what spread you use, though, a huge benefit is that it makes rolls very fast, especially 'roll under, roll high'.

[noob]

d20's evolution from OD&D up to 5e has been interesting, because a lot of the above problems have been ironed out. Both 4e and 5e modifiers and the like where designed intentionally, unlike almost every other RPG engine I've taken apart.

In 4e both the modifiers and the target values are set so that odds of success stays close to being the same over time.
In 5e the modifiers are designed to follow a specific progression but the targets are often undefined(looking at you, skills) or varying a whole lot ending up making the success rate extremely random and depend mostly on the table. At some tables everybody will fail at everything including tying their shoes while the gm have no intent to make the game comedic because the gm just wants players to fail 50% of the time, in some other tables the players will keep succeeding because the gm wants the players to follow the plot which is not possible if the players simply fails to reach the plot and so on: you might as well remove the modifiers and have the gm define the success rates: Not designing the targets for the rolls is the same thing as not designing the modifiers, just remove the modifiers and life will be easier.
4e did fix the issue, 5e reintroduced the issue.

[Yakk]

In 4e both the modifiers and the target values are set so that odds of success stays close to being the same over time.
In 5e the modifiers are designed to follow a specific progression but the targets are often undefined(looking at you, skills) or varying a whole lot ending up making the success rate extremely random and depend mostly on the table. At some tables everybody will fail at everything including tying their shoes while the gm have no intent to make the game comedic because the gm just wants players to fail 50% of the time, in some other tables the players will keep succeeding because the gm wants the players to follow the plot which is not possible if the players simply fails to reach the plot and so on: you might as well remove the modifiers and have the gm define the success rates: Not designing the targets for the rolls is the same thing as not designing the modifiers, just remove the modifiers and life will be easier.
4e did fix the issue, 5e reintroduced the issue.

Task Difficulty DC
Very easy 5
Easy 10
Medium 15
Hard 20
Very hard 25
Nearly impossible 30
If your DM considers "tying your shoes" to be, say, "Very Hard", then yes, success rate is going to be low.

Of course, a T4 character who cares about a skill can pull off DC 30s 50%+ of the time. The current T4 PC I'm playing has a passive perception of 28, for example.

DMs are free to ignore what the rules say for DCs, yes. But the idea that there are no rules is not true.

But really, "tying your shoes" is very easy or easy.

Now, it is true that "walking a tightrope in a storm" could be anything from Medium to Nearly Impossible depending on the DM's interpretation of how hard that is.

[Anonymouswizard]

Many percentile systems end up working in increments of 5 or so, which leads to them effectively having the same spread as a d20 roll, albeit with immediately obvious probability of success and the mentioned tricks you can do with the digits. Regardless of what spread you use, though, a huge benefit is that it makes rolls very fast, especially 'roll under, roll high'.

Even if the system doesn't in games where placement is completely under player control most players default to multiples of 5. It's just easier, and honestly 'I increase my flouncing skill by 5' just feels better than 'I increase my flouncing skill by 1'. Humans like bigger numbers (which I suppose is why Rolemaster used 1d100+modifiers).

Which isn't an issue. Percentile systems have other advantages, and d20 roll under systems work fine. The early percentile systems using 'roll under, roll low' was a big mistake even if it seemed to make more sense, I'm glad it's mostly faded away in newer games in favour of 'roll under, roll high'.

[Old Harry MTX]

You could also think of an hybrid system where you use a Master Die (E.G. 1d20, 1d100 or 2d10), to which you add the number of success from a limited pool of Bonus Dice (ideally d6s), instead of adding a static bonus as usual (+1, +2, etc ...). This system, however, is only interesting for those who like to throw handfuls of dice at each roll.


Sometimes I like to contemplate a hybrid system. If your modifier is zero, you roll one die, and the success or failure of that die determines the results of your roll. If your modifier is positive, you roll modifier + 1 dice, and succeed if any of them come up as successes. If your modifier is negative, you roll abs(modifier) + 1 dice, and only succeed if all of them come up as successes.

Uh, I like it!

[Anonymouswizard]

It's also worth considering things like Savage Worlds, where your stats are dice types. This makes the floor stable while raising the (likely) ceiling, which is very different to how something like D&D works. A high rank SW character is less likely to fail a roll, but it takes a lot of work to get to the point where you can't fail a standard difficulty roll.

To elaborate, in SW a target number of 4 is about equal to DC10 or 15 in D& Depending on the group. It's standard. A D&D character in 5th edition can automatically succeed on a DC10 check in 5th edition by level 9 if they focus on it, a character with expertise can automatically succeed at a DC15 check by level 13. It's just a thing that happens, eventually you don't have to roll (and it happens earlier in 3.X*). Comparatively get a human character up to Legendary rank in Savage Worlds and take everything that boosts your chance and you'll max out at d12+2, which still fails a standard difficulty if you roll a 1 (and <4 on your Wild Die). While SW doesn't have a cap on advancements it's still using endgame abilities to get around the rarity of static bonuses, and even then it isn't certain. Now you can boost it further (via picking a race that gets a bonus to the trait or using the superpower rules), but it requires much more effort and was practically impossible for PCs until the latest edition.

Now dice pools have this behaviour as well, but stats as dice gives the arguable benefit of having a flat probability curve (assuming fair dice). On the downside it in practice caps stats at a d12 unless you begin adding modifiers (although d14s also aren't hard to find, I remember them generally being considered a bit unfair).

You can also rate stats as multiple dice and either take the highest or the total (or count successes I suppose). One system I abandoned had you roll [attribute]d[skill] and total, so a Strength of 3 and a melee of 8 had you roll 3d8. Other systems use pools of mixed dice, and one of mine rates both Attributes and Skills as dice (you roll both and total).

* 2e and earlier used very different skill mechanics (if any), and 4e recommended scaling DCs to the party's level, so I'm leaving them out.

[noob]

If your DM considers "tying your shoes" to be, say, "Very Hard", then yes, success rate is going to be low.

Of course, a T4 character who cares about a skill can pull off DC 30s 50%+ of the time. The current T4 PC I'm playing has a passive perception of 28, for example.

DMs are free to ignore what the rules say for DCs, yes. But the idea that there are no rules is not true.

But really, "tying your shoes" is very easy or easy.

Now, it is true that "walking a tightrope in a storm" could be anything from Medium to Nearly Impossible depending on the DM's interpretation of how hard that is.

Except gms will have extremely varying interpretations of absolutely everything including things that may look utterly mundane to you.
And some gms will just set up the dc so that you have a chance to fail instead of just not rolling in lots of situations where rolls hardly makes any sense (who cares about if you succeed to knock down the door if there is no opponents around to listen to your failed attempts nor time pressure).
Adjectives are not a good scale because everybody have a different visions of adjectives.

[Squark]

Except gms will have extremely varying interpretations of absolutely everything including things that may look utterly mundane to you.
And some gms will just set up the dc so that you have a chance to fail instead of just not rolling in lots of situations where rolls hardly makes any sense (who cares about if you succeed to knock down the door if there is no opponents around to listen to your failed attempts nor time pressure).
Adjectives are not a good scale because everybody have a different visions of adjectives.

A game cannot cover all scenarios and provide DCs or their equivalent for everything. Adjectives are necessary to provide a GM some sort of sense of scale. I cannot think of a game system I've played that has not provided such a breakdown for GMs, from FFG's percentile based 40k system, FFG's dice pool based Star Wars System, 3.5 D&D and onward, or FATE (Which is essentially a 4d3-8 system abstracted into a plus sign, blank, and a minus sign).

However, one thing that helps ground the ambiguity of such systems are benchmarks. Examples of what is a "very easy," "easy," etc. tasks would make it a lot easier for GMs to ground their understanding of what the developers thought, and that is one thing I feel 5e fails.

As far as my thoughts on the systems;

-The d20 mechanic is easy to understand, but a single die is inherently more random than multiple dice.
-Good D100 systems are just d20 systems with a different skin. The granularity of a few points isn't very useful. You use a D100 system either to distinguish yourself from d20 systems, or because you can the 2 digits for some other purpose (FFG's 40k system used the first digit of each ability score to get a characeristic (E.G. A Space Marine with 43 Toughness has a Toughness characteristic of 4), which they used for things like damage and health, while the second digit in attack rolls was used to determine what body part was hit for critical hits and armor with different levels of protection).
-Dice Pool Systems can have a ton of moving parts or only a few. Fate uses a simple range of -4 to +4 that you add to your base skill and then compare to the target difficulty (which can be further modified by the system's FATE points to either add 2, re-roll your dice pool, or both). This means you have a clear bell curve as to what the most common results will be. On the other end of things, you have systems like the one FFG used for their Star Wars system (I believe they adapted it for their generic genesys system), where you have pools made up of d6s, d8s, and d12s, as well as different symbols on the dice. I found that the system led to some interesting narrative effects, but it was a bit more complicated to teach.