How to write no rules

[Murtak]

PhoneLobster: I have no clue how I could make my meaning clearer to you than it already is. To me it looks like take single words I said, then recombine then into random phrases. I guess it's possible that I am posting compelte gibberish but I certainly can't spot it.

[mean_liar]

Yep.

Also the "too impossible to try" problem with a d20 is dismissed with the 1 and 20 auto-success/failure results.

I think ultimately the difference is between how predictable you want your results to be, since alterations to curved systems create assured results at lower magnitudes.

...

EDIT: PhoneLobster is a very aggressive poster who doesn't consistently communicate clearly but when he does he usually has something to contribute. He doesn't put up examples until the third or fourth page of insults if you can last it out.

[virgil]

...liberal arts majors who took Math for Dummies 101 for their gen-ed requirement. Since those are the people who would most be attracted to rules-lite systems, 3d6 is probably a bad idea for those games.

Didn't oWoD have a similar concern in its resolution system, obtuse probability? When the system can't be fully realized by everyone, it's easier for everyone to go Magic Tea Party while not caring about optimization.

That's another thing. When there is the ability for optimization, anti-munchkins will avoid doing it, no matter how easy it is for them. When the pro-optimizer can't master the system, he's on about the same playing field even after trying. Having that kind of set-up is easier to write-up and design than trying to delicately balance & playtest slight numerical tweaks.

[Murtak]

Also the "too impossible to try" problem with a d20 is dismissed with the 1 and 20 auto-success/failure results.

That gives you scenarios where the lich-king's evil ritual leaves 1 in 20 peasants alive though. Presumably you do want to cut off the RNG in most systems, or least have it taper out into obscurity. 5% is still noticeable for some effects.

[PhoneLobster]

Am I getting this right?

Pretty reasonable. Not keen on the multiplying by fractions thing for obvious reasons very similar to my protests against the "curvy" methods in the first place.

With Murtak descending into "I wants lots of smooth steps, no wait, who the fuck says I wants fine granularity?" territory it's nice to see someone somewhat "getting it".

[mean_liar]

That's acceptable for some and easily overcome with houserules.

For DnD, I think the first thing I thought of is that the 1/20 rule can only be used by creatures with PC classes or race-derived Hit Dice.

Given that there's usually some kind of system difference between PCs and NPCs a similar rule could cover a lot of ground in a lot of systems.

[Roy]

Also the "too impossible to try" problem with a d20 is dismissed with the 1 and 20 auto-success/failure results.

That gives you scenarios where the lich-king's evil ritual leaves 1 in 20 peasants alive though. Presumably you do want to cut off the RNG in most systems, or least have it taper out into obscurity. 5% is still noticeable for some effects.

Well, usually death magic still does some damage even if you pass. And while 3d6 + CL or whatever isn't a big deal for real characters, it is very much capable of dealing with low level NPCs.

[Murtak]

That's another thing. When there is the ability for optimization, anti-munchkins will avoid doing it, no matter how easy it is for them. When the pro-optimizer can't master the system, he's on about the same playing field even after trying. Having that kind of set-up is easier to write-up and design than trying to delicately balance & playtest slight numerical tweaks.

Even if you could do this you will end with some random group members being less/more powerful than the others, for no discernible reason. That stinks just as much as a system where the optimizers win by default.

[PhoneLobster]

When the pro-optimizer can't master the system, he's on about the same playing field even after trying. Having that kind of set-up is easier to write-up and design than trying to delicately balance & playtest slight numerical tweaks.

Except despite what "real role players" claim they too aim for success.

Making the system hard to "game" doesn't stop people from "optimizing". It just means that those few who can will have a significantly larger margin of success and imbalance compared to those who can't. And those who can't won't know why or how. It will look to them like cheating or black magic or GM favoritism or something.

As for "critical"/auto success and failures on the extremes of dice rolls. I'm largely against them but it seriously isn't an issue in this thread yet, and doesn't really need to be.

Indeed considering the discussion I posted about how arbitrary "Yes" and "No" rulings from the GM are a vital part of doing rules lite stuff well I'm not sure if there is really a place in that for "anything succeeds on a 20".

[MartinHarper]

That WILL require something other than a flat d20. But it requires a very, very specific "curve". And NOT one of the ones already mentioned.

Specifically, you'd need an exponential decay curve where P(x) = A^x. That way, if you add +1 to the target difficulty, you reduce the chance of success by A. So you could do this with rules like:

1. Roll a number of d6 equal to the target difficulty.
2. If they all come up 3+, you win. Gratz!

However, the tail of a 3d6 curve is probably close enough to an exponential decay for gaming purposes, and is certainly closer than a 1d20 'curve'. So you're weighing transparency against the desire for exponential decay, and coming up with a compromise in the form of 3d6.

[PoliteNewb]

Okay, I'm going to try to clarify this...since I think I just got what Murtak is talking about, and it's apples to your oranges, PL.

No. I don't care about granularity,

But you said you wanted a larger number of finer sized bonuses!

Actually, no...he said he wanted to be able to add more bonuses without going off the RNG. He doesn't care EXACTLY how fine they are, as long as they don't push us off.
Basically, he doesn't care if the flat bonus to the die roll changes the chance of success by a little or a lot for any particular die roll. All he cares about is what chance it alters in an average situation, since those are the most likely to occur. If I'm reading him right, it is good to have that bonus mean one thing most of the time, but still be able to apply it multiple times without fucking up the situation too badly.

And I don't want a flat RNG, because it fits my goals worse.

But you said you never want to recalculate the value of bonuses!

He doesn't. He likewise doesn't care what fucking effect the bonuses have on the ends, as long as the effect is not "whoops, off the RNG motherfucker".
All he cares about is how the bonus affects an average situation, and whether or not he can continue to apply it without pushing off the RNG completely.

Again, no. I want to halve the chance for the common case.

But you said you don't want to recalculate the value of bonuses. If you pick a point in a curvy RNG to define your bonus then fix the size of the bonus, it's value fucking changes.

We're aware. We also don't care.
The whole point of this is that it's OKAY for bonuses to have different meanings. That's why we don't want a flat RNG. You keep pointing out that he's arbitrarily picking a point in the middle...YES. Because on a curved RNG, the middle is what matters most of the time.

On a flat RNG, a bonus will mean the same thing wherever you are...and stacking bonuses will rapidly push you off.
On a curved RNG, a bonus will mean different things depending on where you are, but stacking bonuses will gradually become less and less likely to push you off.

(Note that this does pretty much prove PL's point about transparency/opacity...if a bonus to a 3d6 roll means different things depending on where you are, that is fairly opaque to someone who doesn't do probabilities well...like me, for instance)

Murtak (and Frank, I guess), hope I didn't misrepresent you. Also hope this clears up why the fuck people are discussing different things.

Long story short:

Curve people want a bonus/penalty to have an average meaning. Yes, this means that in edge cases the bonus/penalty does not actually mean what you think it means, but it does mean you can just toss bonuses/penalties on without having to recalculate AND without being pushed off the RNG.

PL, I believe you think that pushing the chance of success so far into the corner is the same as pushing it off the RNG. Others disagree.

[virgil]

I know in FUDGE's bell curve dice scenario, situational modifiers are not supposed to be an aggregate of bonuses/penalties, but usually just the biggest (once you're blind, the fact it's raining and they have higher ground doesn't mean anything). I've seen some discussions where multiple equal-sized modifiers can sum up to a single larger one, but it's never in a 1-for-1 scenario; such as two or more -2 penalties mean you have a single -3.

The idea is that because of the bell curve, modifiers past the initial +/-2 are drastic, and you absolutely cannot treat them as simple arithmetical modifiers.

[Ice9]

Essentially, it seems like what some people really want is a logarithmic RNG, where you would never actually reach 0% or 100% - just get increasingly closer. But since that would be hard to simulate via dice, a bell curve is closer to it than a flat RNG is.


Although incidentally, there is a (somewhat slow) way to simulate it with dice - roll an extra die for each "step" of your logarithm, each die having a fixed chance of success. Somewhat like a dice-pool system, but instead of getting a fixed number of successes, you succeed if any dice are a success (for positive modifiers), or if all your dice are a success (for negative modifiers).


Edit: MartinHarper ninja'd me (or maybe I wasn't paying attention).
Yeah, that example is pretty much what I meant, with the exception that positive modifiers can make it go logarithmically easier instead of harder.

[MartinHarper]

Yeah, that example is pretty much what I meant, with the exception that positive modifiers can make it go logarithmically easier instead of harder.

Heh. That way is much clearer.
I wonder if that's where dice pool mechanics came from? Someone wanted a logarithmic RNG, so they made "roll Xd6, get at least one 5+", and someone else thought this wasn't complicated enough, so they started having variable target numbers and variable dice size and variable numbers of required successes.

[PhoneLobster]

However, the tail of a 3d6 curve is probably close enough

"Close Enough" doesn't meet Murtaks "No recalculation" requirement.

Actually, no...he said he wanted to be able to add more bonuses without going off the RNG. He doesn't care EXACTLY how fine they are, as long as they don't push us off.

He specifically stated "more and smoother".

And yes he doesn't care, or particularly understand, that the flat bonus has wildly variable effects on chances of success.

As for the bonus meaning "one thing most of the time" when is that? changes and drop offs on 3d6 are pretty dramatic. A +2 bonus can be worth less than 2% or more than 20% that's more than a magnitude of 10. If that isn't "fucking the situation up too badly" then what is?

All he cares about is how the bonus affects an average situation

Which he isn't using in his examples The calculations he is making depends not on the average value of the bonus but on the value of the bonus at a specific point in the graph.

That is fundamentally bad math.

On a flat RNG, a bonus will mean the same thing wherever you are...and stacking bonuses will rapidly push you off.
On a curved RNG, a bonus will mean different things depending on where you are, but stacking bonuses will gradually become less and less likely to push you off.

This is incorrect.

You are making the mistake that Murtak's bad math is trying to cover for.

What we are discussing is basically a situation where we have 10cm in which to draw a graph covering values between 100 and 0. We can only actually read values at whole 1cm increments of the 10cm space we have to draw the graph in.

One faction says "Hey!" If we make the graph, logarithmic, and the first few centimetres cover 50 of the 100 points of value and the last few cover only parts of 1 point of value, then why the number of whole 1cm increments we can read will increase.

This is incredibly lame and incorrect. Look back at Murtak's 1d11+1 vs 2d6 example for the best demonstration of this. Both are ranges of 2-12, and yet he claims you can add more bonuses to one of them without leaving the range!.

Do you see where that is complete and utter bullshit? 2+2=5 territory?

Our bonuses are a limited range of whole numbers a "curved" mechanic in practice does not approach zero as it approaches infinity, it approaches zero as your bonus approaches (in great lurching whole number steps) it's maximum range. It just drops off the edge of our limited range of whole number dice results the exact same way a flat d20 does.

The only difference is that before it did that several of the last increments of that range were statistically worthless.

but it does mean you can just toss bonuses/penalties on without having to recalculate AND without being pushed off the RNG.

Where do you get this idea from?

The calculation used to set the size of the bonus that Murtak and Frank presented wasn't an average calculation.

And it also does NOT meet Murtak's "half chances each time" scenario.

We are working with a limited linear range in which to add bonuses you WILL be pushed of the RNG if you just keep adding bonuses. "Curves" do not stop that. Curves with average bonus calculations do not even begin to impact on that.

And hell yes 0.4% is a worthless increment on a scale with various 10-15% sized increments. So I sure as heck think that you are pushed off the "functional" RNG in those scenarios before being pushed of the actual complete RNG. Which you will be, because the number of whole number bonuses available is not magically infinite or even any larger than what is available on a flat RNG.

[Username17]

Oh crap. Walked away for a day and PhoneLobster took a crap all over his own thread.

PhoneLobster: holy crap it is difficult to try to have a discussion with you. I laid down some rather simple math in defense of a mechanic that is solid, well used, well liked, and not even a mechanic that I personally use very often. And you... used some atrocious math, willfully misunderstood everything that everyone was trying to say to you, and layered insult after insult on people who were specifically and disinterestedly trying to help. Fuck off dude. Stop being such a raging asshole. That's coming from me. Got that? Frank Trollman thinks that you're too much of a dick in conversation and debate. It's offputting.

For the last fucking time:

• The effects of any numeric shift will vary depending on where you are on the RNG to begin with. A +1 on a d20 will double your DPS if you naturally hit on a 20 and it will add 10% to your DPS if you hit on an 11+. On a curved die roll, that change is bigger, just as the change becomes bigger if you use a larger die (100% to 2% DPS shift from the end to the middle on percentile dice). Some people regard that as a good thing, but it is in no way a unique thing.
• Your game is going to have a list of modifiers. Maybe you're looking on page 151 of the 3.5 PHB. Maybe you're looking on page 140 of SR4. I don't even care. You're going to have a number of modifiers. On a curved die roll, you can set any one of those modifiers to a value that you can feel and notice by itself, and yet still be able to linearly add more of those modifiers before falling off the RNG entirely than you can with a flat roll.

These are just facts. If you don't accept them as facts, you're a douchebag. Period. They are not reason to abandon uncurved die rolls as a core mechanic for all games, but they are reason enough that many people who want to use curved die rolls for stuff are not wrong to do so.

But PL, you are wrong. Mathematically you're wrong. The fact that you're being an uncompromising asshole about it doesn't help, but the important part is that you're wrong. It was actually a relatively minor point in your initial tirade, but you have allowed it to explode and poison your entire thread and now your thread serves no purpose save as a brutal indictment that you are an uncompromising dick who can't learn to shut up when he's dug himself a hyperbole hole big enough to see his unwashed cock through.

-Username17

[PhoneLobster]

And you continue to use your basic falsehood about how two ranges of the same whole number bonuses are magically differently "deeper" because... you say so.

You continue to define "a bonus you will notice" by it's value at one point in the graph.

You are a liar and an idiot Frank. Your work in this field increasingly is nothing but worthless, worse than that, deliberately misleading.

No matter how much shit you spin.

100% of one 0-100% range does NOT have more 30% sized increments than another 0-100% range.

A 3-18 whole number range does not have more discrete whole numbers in it than a 1-20 range.

And converting a 30% increment into a whole number bonus on a 3-18 range does not give a larger number of 30% increments on that range OR a larger number of whole number bonus increments on that range!

You are literally pulling numbers OUT OF YOUR FUCKING ARSE.

[Username17]

It's not "one point on the graph" it's THE FUCKING STARTING POINT ON THE GOD DAMN DIE ROLL YOU RETARDED, DISHONEST ASS HAT!

There's a base Target Number you need to hit or to succeed or whatever. It's usually 50-60% in most games. All modifiers adjust that number. When there's only one modifier, it can be big enough to notice or big enough to dominate notice depending on its size and the type of random number generator. On a curved die roll, more such modifiers can apply without pushing the TN entirely off the RNG one way or the other. That is a fact. It's a fact that is useful and important to many people when they are designing games.

And I really think you understand that. I really believe that you're just lying to people to try to win points on me or something. And yeah, now I'm pissed off at you. Again. Increasingly of late your "contributions" to this board have become more shrill and more insane. Honestly I don't give a rat's ass about your rants any more because they aren't fun. They are just you insulting people like a guy with mouth diarrhea.

Seriously: Fuck off and die in a fire. My patience with your bullshit is exhausted.

-Username17

[PhoneLobster]

On a curved die roll, more such modifiers can apply without pushing the TN entirely off the RNG one way or the other.

You don't get to add +30% more often.

That is a fact. It's a fact that is useful and important to many people when they are designing games.

As long as they can't or won't do consistent math.

I really believe that you're just lying to people to try to win points on me or something.

That would be projection.

[Username17]

Sigh. That's that.

Phone Lobster: you are on ignore now. You have ceased being useful.

-Username17

[PhoneLobster]

You have ceased being useful.

Again. Projection.

edit: And for those that aren't flipping the ignorant switch because they got their panties in a twist for being called out on an inconsistency. Again.

I would like to briefly address Frank's raising of bonus effect on DPS rather than on an individual rolls odds of success.

He is effectively adding an additional factor into consideration, a new axes of "Time" over which multiple rolls are made. And thus showing that bonuses have a different and more complex effect (over time) depending on their position on the range.

Hey, you know I can play that game too. You know that the results of using a flat d20 roll to determine success ALSO become even more opaque, complex and non linear if I require that the d20 and the 3d6 make second rolls of the same type, like a damage or soak roll, in order to be successful! Thus proving the flat d20 to be non linear in nature! Oh and if we measure that over time AND apply a factor that changes the bonus to each by some arbitrary rate over time then the d20 roll becomes even more complex and non linear!

I hope you can all see the utter contempt this sort of false comparison has for it's audience. Watch as my right hand piles on more and more additional multipliers, variables and factors to the d20 mechanic, while my left hand just as quickly ignores and stuffs the effects of those same factors on a 3d6 under the rug.

The fact is the discussion is supposed to be primarily about odds of success on a single roll. For use in an utterly skeletal "roll to succeed or fail" rules lite type of system. And even beyond that should you pile on complex multipliers to your basic mechanic, starting with a more transparent and functional mechanic is still productive as long as you apply all your modifiers evenly to both sides of your comparison.

[User]

I would like to briefly address Frank's raising of bonus effect on DPS rather than on an individual rolls odds of success.

Frank is wrong in this case, but not for the reasons you cite. Let's take the DPS case. Assume your average damage is 20. The following chart gives average damage if you hit on an X or higher

X: DPS
20 1
19 2
18 3
17 4
16 5
15 6
. .
. .
. .
3 18
2 19
1 20

As you can see your DPS goes up by 1 for each +1 bonus to hit you get, anywhere in the scale, contrary to what Frank said. If you hit on a 5 or higher and get a +2 to hit you get +2 to your DPS. And if you hit on a 10 or better and get +2 to hit you get +2 to your DPS. The fact that there are repeated rolls being averaged out here is not at issue.

Frank's error is not mathematical. It is true that if you apply a constant bonus to two different numbers the percentage changed will be different if the original values are different. 1+1=2, a 100% increase; 10+1=11, a 10% increase. Frank's error was simply that this is not what was being talked about when the difference between a bonus giving a constant increase or a variable one was brought up. The difference was, for example, between the fact that a +1 on a d20 always affects your chances by 5% if you are on the RNG, and that a +1 on 3d6 can affect your chances by anywhere from 10% to 1% (roughly) depending on what number you are trying to hit. Apparently Frank either didn't realize, or chose to ignore, that what was in question was the size of the change in your chances given a bonus, not the size of those changes proportional to what your chances were previously. (Or with damage, the change in your DPS, not the change proportional to your DPS before the bonus.)

[ckafrica]

The problem with the bell curve is that the further away from the mean you get, the less likely it is to come up. This causes modifiers to mean increasingly less the further from the mean they push you.

Let's assume a base target of 10 or less on 3d6 (like gurps) which is a 50% chance of success. Now a +1 increases you chance of success by 12.5% while a -1 reduces your chance by 11.57%.

At base target 11 the value of those modifiers reverse +1 gives you +11.57% while -1 is -12.5%

Now if by base target was 13 I'd be starting at an 83.79 percent chance of success. A +2 will increase by chance of success by only 11.66 percent while a -2 will reduce it by 21.3%

If we go to a base of 15 (95.37%) a +3 only gives us a +4.63% while a -3 is -21.2%

The effect of using a Gaussian curve rather than a set rate of probability is that the effect of all your actual modifiers is different depending on your starting point. A -3 can actually be less detrimental to a a high base target than a -2 to a lower base target (this of course is until you go below 11 at which point it switches)

What the Gaussian curve does in game terms is make strong characters less affected by modifiers than mediocre characters. It also gives you a diminished return moving stats off the median

So PL is right to the extent that you can't say a +4 on a d20 should be a +2 on 3d6 because they won't result in the same effects. The 3d6 creates a completely different set of probabilities and so modifiers against can never replicate the effect of modifiers on a d20

[PhoneLobster]

Excellent ckafrica, I give you a pass in high school maths. Carrying all the authority of me, some guy on the internet.

Frank is wrong in this case, but not for the reasons you cite. Let's take the DPS case.

When Frank was talking about DPS he was basically talking about the rate of change of a +1 bonus, proportional to the "bonus" or TN that already existed.

Or in other words he very informatively told us. "2 is twice as much as 1, and 1 is one tenth of 10!" Very interesting I'm sure but largely irrelevant to anything.

Those claims both hold in regards to your DPS example if you look closely. It's just that there is no actual connection between those claims and actual useful information, certainly no connection between those claims and say, the actual chance of success or actual damage output of a given roll.

Basically someone was talking about how fast a car was moving and Frank chose to go off and discuss the proportional acceleration in order to avoid the fact that the car is doing over 60.

[Username17]

The problem with the bell curve is that the further away from the mean you get, the less likely it is to come up. This causes modifiers to mean increasingly less the further from the mean they push you.

How is that a problem with Bell Curves? As the bonus gets larger they mean less, but they stay on the RNG. At least, they stay on the RNG longer. Your bonus that increases your odds by 25% will, when added a second time, add less than 25%. That's not a problem, that's the entire fucking point, since with a linear number generator we'd have already left the game entirely and gone to the realm of magical teaparty declarations.

If the argument in favor of something is "This does X!" then it's pretty vapid to respond to that with "Oh yeah?! Well it does X! What do you say to that?!" Yes, a curved number generator by definition causes a 10% reduction to be itself reduced when applied iteratively. It does this and therefore your 10% reduction can be applied more times.

Frank is wrong in this case, but not for the reasons you cite. Let's take the DPS case.

Good point. DPS is the wrong word, or at least could easily be construed as such, making it a poor choice on my part. A linear number generation on a to-hit roll does add an equal total amount of average damage for each bonus no matter how many bonuses you have or do not have. Mea Culpa if I implied anything different.

However the point I was trying to make still stands, which is that actual offensive effectiveness changes more noticeably at the edges than at the middle. In D&D, which I remind everyone is an entirely linear number generation system, AC is worth more the more you have. This is because the actual game is that you lose after you have been hit a certain number of times, and when you go from being hit on a 19 to being hit on a 20 that single +1 to your AC literally doubles the amount of time you stay up. It's roughly equivalent to doing twice as much damage for purposes of whether you win the fight or not.

So yes, the average damage inflicted per unit time does change linearly, the value of that linear change varies wildly. And it varies more wildly the finer you make your random number generator.

-Username17