Homebrew 2d10 vs 1d20

[leozg]

Bell shaped curves (2d10, 3d6...) are good for non opposed rolls, where one side is variable (your skill/proficiency/ability to achieve the result) and the other (dificulty) is fixed. Since the variable side has a low variance/dispersion (when you're good at something that only depends on you, you're likely to succeed with regular frequency), the results are more predictable. Good for ability tests (non opposed rolls). The "take 10" rule goes in thia way, but it reduces the variance to 0, too drastic.
For tests where both sides are variable, like attacks or other opposed rolls, a linear result (1d20) is better since your skill is being tested against someone's else skill, so you want the results to be less predictable.
So if someone is open to abandon the d&d "1 roll to rule them all" rule from since 3e, I would recommend to use both rolls.

 

[SableWyvern]

Bell shaped curves (2d10, 3d6...) are good for non opposed rolls, where one side is variable (your skill/proficiency/ability to achieve the result) and the other (dificulty) is fixed. Since the variable side has a low variance/dispersion (when you're good at something that only depends on you, you're likely to succeed with regular frequency), the results are more predictable.

The results on the dice are more predictable. Whether the result of the skill check is more predictable depends entirely on the system built around the dice and the what the range of the variables is.

If the target is 8+ on 1d10 (fixed difficulty), basic proficiency provides +3, advanced proficiency gives +5 and expert gives +6, then advanced and expert characters succeed 80% and 90% of the time respectively. That's very predictable. Unskilled people fail 70% of the time, which is also quite predictable. You don't need a bell curve to have predictable outcomes.

 

[NotAYakk]

The results on the dice are more predictable. Whether the result of the skill check is more predictable depends entirely on the system built around the dice and the what the range of the variables is.

If the target is 8+ on 1d10 (fixed difficulty), basic proficiency provides +3, advanced proficiency gives +5 and expert gives +6, then advanced and expert characters succeed 80% and 90% of the time respectively. That's very predictable. Unskilled people fail 70% of the time, which is also quite predictable. You don't need a bell curve to have predictable outcomes.

I think your definition of predictable isn't what other people are talking about. "I as the person running the game can calculate the chance" being "predictable" isn't a super useful definition of "predictable", and definitely isn't the one being used here.

You could talk about being able predict how hard a challenge would be to a party when designing it. But you can also talk about "when a character tries to jump 10', how predictable if they make it"? A 0% chance and 100% chance there is predictable (always fail, always succeed), while a 50% chance (even if you know it is 50%) is in a sense "this is an unpredictable result".

How predictable the odds are vs how predictable the result of the task is, if that makes sense.

Careful about semantic arguments like this. They don't go far usually.

 

[SableWyvern]

I think your definition of predictable isn't what other people are talking about. "I as the person running the game can calculate the chance" being "predictable" isn't a super useful definition of "predictable", and definitely isn't the one being used here.

You could talk about being able predict how hard a challenge would be to a party when designing it. But you can also talk about "when a character tries to jump 10', how predictable if they make it"? A 0% chance and 100% chance there is predictable (always fail, always succeed), while a 50% chance (even if you know it is 50%) is in a sense "this is an unpredictable result".

I'm really not sure what you're saying here.

Being able to know in advance how likely you are to succeed is very important when it comes to making decisions in character. Knowing how difficult a challenge is likely to be when planning an adventure is typically very useful to a GM. You seem to be suggesting this is not the case, which does not at all align with my experience.

I absolutely agree that 100% and 0% chances are the most predictable, while 50% chances are the least predictable/most swingy. The aligns entirely with my point.

How predictable the odds are vs how predictable the result of the task is, if that makes sense.

I genuinely do not understand the distinction you're drawing here. The odds tell you how predictable the result of the task is. These are the same thing, to me. I note that you, too, used the odds in your example, in exactly the same way that I did in mine.

I used the phrase, "the result of the skill check". To me, this is entirely synonymous with "the result of the task", on the assumption that the output of the skill check tells us the result of the task.

Careful about semantic arguments like this. They don't go far usually.

Again, I really don't understand what you're trying to warn me about here. I'm happy to work with any definition of "predictable" people want but, while it's clear you don't like the definition I'm using, I do not have any idea how your definition differs.

In the post you quoted, I provided an actual rule example of a 1d10 skill system and claimed it provides predictable results. Do you disagree with this? If so, can you explain why it doesn't meet your definition of predictable?

Edit: It occurs to me that maybe you thought I was arguing that it's harder to recognise odds when you roll 3d6 (eg, most people will have very little idea what the actual chance is of rolling 14+ on 3d6). While it is certainly the case that recognising the actual odds is harder, and is one reason why bell-curve systems can sometimes make it harder for people to accurately estimate their chances of success, this it is not at all the point I was trying to make. For the purposes of the argument I was making, you are free to assume the participants are perfectly and intuitively capable of recognising that the chance of rolling 14+ on 3d6 is 16.2%, and that bell curves make this type of estimation no harder than any other system.

 

[leozg]

You don't need a bell curve to have predictable outcomes.

For a specific roll with a declared difficulty, you can calculate the odds, but that's not what I meant by predictable. My analysis is focused on the roll result you can achive and the variance of this result, DC comes later.
Let's consider a long jump Olympic athlete. This guy knows in advance how long he'll jump,, with a small error margin, because he's being training this for months/years. This is low variance. In D&D if how long this guy will jump is a roll result, for sure a flat roll like 1d20 is the worst mechanic you can model to this situation and a bell shaped curve is better.
I recommend 2 readings on this matter:
- Torben's article on dice mechanics:

https://www.darkshire.net/jhkim/rpg/systemdesign/torben_rpg_dice.pdf

- A Treatise on Different Dice-rolling Mechanics in RPGs:

A Treatise on Different Dice-rolling Mechanics in RPGs - RPG Design

rpg-design.wikidot.com

 

[SableWyvern]

For a specific roll with a declared difficulty, you can calculate the odds, but that's not what I meant by predictable. My analysis is focused on the roll result you can achive and the variance of this result, DC comes later.
Let's consider a long jump Olympic athlete. This guy knows in advance how long he'll jump,, with a small error margin, because he's being training this for months/years. This is low variance. In D&D if how long this guy will jump is a roll result, for sure a flat roll like 1d20 is the worst mechanic you can model to this situation and a bell shaped curve is better.

"In D&D" is far more important to the point you're making than "1d20".

It's not the use of the 1d20 that creates the problem you're pointing it, it's the entire system, of which the d20 is just one part.

If I'm using 1d4 (flat) for my long jump check, I know my result will lie within a range of 4 units. If I'm using a 3d6 (bell curve), with the same value for the units, the distance I will jump suddenly includes results over a vastly wider range.

I recommend 2 readings on this matter:
- Torben's article on dice mechanics:

https://www.darkshire.net/jhkim/rpg/systemdesign/torben_rpg_dice.pdf

This is far too much to read, but at a brief skim I didn't see anything in it that's suggests it's going to disagree with my point. If you would like to point out exactly where it does so, I'll be happy to read closer.

- A Treatise on Different Dice-rolling Mechanics in RPGs:

A Treatise on Different Dice-rolling Mechanics in RPGs - RPG Design

rpg-design.wikidot.com

This person clearly recognises that you need to analyse systems, not just the dice in isolation, which is exactly the point I'm trying to make.


Using 3d6 doesn't magically give you a better or more realistic range of options. The effect that 3d6 roll has in your game depends entirely on the system built around it.

Using 1d20 doesn't magically give you inferior or a less realistic range of options. The effect that the 1d20 roll has in your game depends entirely on the system built around it.

There are many people out there who think dice probability begins and ends with statements like "using 3d6 for skills is more realistic" or "one benefit of choosing 3d6 for your resolution mechanic is that players have more control over outcomes", and this is what I'm fighting against. You can't assess a system, or make meaningful statements about that system, or give advice on how a system will work, based on the dice alone -- the system needs to be taken and assessed as a whole.

Edit: And, for the record, using 3d6 seems like a much saner starting point than 1d4 if you're trying to design a system for long jumping. My point is that choosing 3d6 isn't going to make your system automatically handle long jumps better than a well put together system that uses 1d20. You are still going to have to care about outputting realistic results and put in work and effort into the entire system, and maybe even make compromises elsewhere, to get that long jump system humming.