PDA

View Full Version : Different resolution mechanics in the same system



ThiagoMartell
2012-12-16, 01:55 AM
In old games, we used to have several different resolution mechanics in the same system. In AD&D, sometimes you had to roll low, sometimes you had to roll high, for example. Thief skills used d100%. It was a mess.

The brazilian system Daemon suffers from similar problems. With Daemon it's even more egregious, since the "system" is just a bunch of stuff from other systems copy-and-pasted. There is absolutely no internal consistency. Can't believe those guys were not sued, actually.

However... is this really such a bad thing? Using different resolution mechanics gives a different "few" for every part of the game. I know it just felt more special using Thief skills in AD&D than it did using Stealth in 4e.

Recently I've been designing my game and I have one mechanic for regular tests (roll x dice, matching numbers are successes, somewhat similar ORE) and a different mechanic for combat (roll 2d6+modifiers to attack). The reason is twofold - I wanted a different "feel" for combat and I wanted the rolls in combat to be under a bell curve.

So, all that said... are different resolution mechanics always bad? Any examples of when they are bad (or not)?

warty goblin
2012-12-16, 02:32 AM
Unwarranted Statistical Advise:
If you're just using attack roles to determine success (and not damage bonuses or anything like that) there's no really good reason to worry about a bell curve. If your probability of success is 1/2, it doesn't matter whether your generating a normal random number and seeing if it's above the median, or flipping a coin and seeing if it's heads. The distribution of success and failure is exactly the same. Percentile dice make your probability of success a lot easier to calculate, and allow a lot more granularity than 2d6. Since the distribution is the same, why not use them?

Damage rolls are where having a bellcurve actually makes a difference, since there the number you roll matters, not simply whether it's larger than another number. A weapon that does 10d6/10 rounded to the nearest integer for example is far more predictable than one that does 1d6.

Flavorwise, I'm all for having different mechanics for different tasks. It's one reason why I like systems that use dice sizes as skill levels, different tasks with the same character literally feel different. They suffer from some annoying things to do with variance though. An idea I've been playing with is allowing a person to buy 'expertise' in a particular skill by swapping a d8 skill dice for 2d4. I'd have to sit down with some paper and hammer out how much of a difference it really makes though.

Fhaolan
2012-12-16, 02:32 AM
I once ran a time/dimension spanning game where each character was from a different ruleset. One was AD&D, one was GURPs, one was TMNT (Paladium), and the last was ... Prime Directive, I think?

Not recommend if your players are rule-lawyer-types, but since my campaigns aren't combat-focused as a rule it worked out for me.

ThiagoMartell
2012-12-16, 03:14 AM
Unwarranted Statistical Advise:
If you're just using attack roles to determine success (and not damage bonuses or anything like that) there's no really good reason to worry about a bell curve. If your probability of success is 1/2, it doesn't matter whether your generating a normal random number and seeing if it's above the median, or flipping a coin and seeing if it's heads. The distribution of success and failure is exactly the same. Percentile dice make your probability of success a lot easier to calculate, and allow a lot more granularity than 2d6. Since the distribution is the same, why not use them?
But why use them? Percentile dice don't match my system and they generate big numbers. By using 2d6, I can make sure about 50% of a hit is luck and about 50% is skill (for an average person, against an average person) and that the difference luck makes goes down as you become more skilled.
Attack rolls in my system are made by rolling 2d6, adding two stats (they go from 1-6 each) and comparing it to the target's Defense score. I could change that to use percentile dice, but I would have to multiply scores or change the range of scores. Not very helpful.


Damage rolls are where having a bellcurve actually makes a difference, since there the number you roll matters, not simply whether it's larger than another number. A weapon that does 10d6/10 rounded to the nearest integer for example is far more predictable than one that does 1d6.
In my game, critical hits come up when you beat a target's Defense by a margin of 5 or more. So the number you roll matters.


Flavorwise, I'm all for having different mechanics for different tasks. It's one reason why I like systems that use dice sizes as skill levels, different tasks with the same character literally feel different. They suffer from some annoying things to do with variance though. An idea I've been playing with is allowing a person to buy 'expertise' in a particular skill by swapping a d8 skill dice for 2d4. I'd have to sit down with some paper and hammer out how much of a difference it really makes though.
Looks like a very small difference from where I'm standing, since the average of 2d4 is only higher than the average of 1d8 by 0.5, but I'm sure I'm overlooking something.

Arbane
2012-12-16, 03:27 AM
Having different resolution mechanics CAN work, if you're using it for flavor reasons. Consider a modern game where a melee attack roll has to beat a defense roll, but a gun attack roll just has to beat a target number. (You can't really dodge bullets...)

erikun
2012-12-16, 03:49 AM
The problem with older systems is that, while they used different mechanics because each one worked to model what it was intended to, using them all together made the system feel like it was put together with hot glue and twine. You had to learn an entirely new system for a relatively minor part of the game, which was frequently completely strange or counter to other parts of the game.

That doesn't mean that systems with different mechanics are a bad thing, though. As Arbane points out, it can work quite well if your different-mechanic is representing something special. (I would also recommend no more than one or two "different" mechanics.)

You also want to think of how your two systems would interact with each other, when the time comes. For example, what if one character is trying to attack with a bow and another trying to defend by maneuvering a horse? What happens when someone tries to defend against a character setting off a trap or explosives? If your two different systems are too out of sync with each other, then you'll run into situations where one character arbritrarily fails because they are forced to use/to not use combat stats.

TheOOB
2012-12-16, 04:12 AM
I general, you want to have as few resolution mechanics as possible. The most important thing to remember about an RPG is that it is a game, and game mechanics should always come first when designing an game, as it is far easier to mold fluff around mechanics than the other way around.

The thing is, people need to remember how to play your game. Every time the players need to open the book to reference something during play, the game rules probably failed them. In fact, when designing systems, I like to follow the 2 drink rule, if you can't remember the rule after drinking two beers, it's two obscure/complex.

There are occasions where you may want another resolution mechanic, but those occasions are rare. If you're trying to do X, and you mechanic doesn't handle X well, you have to ask yourself an important question, how essential is X to the game system as a whole. If it's not very essential, just mold X to fit existing mechanics, and accept the fact that your system isn't best at everything. I don't play Paranoia for deep tactical combat, and I don't play D&D because of it's in depth rules for terror and mental derangements.

If X is essential, maybe your mechanic needs to be changed to something that works with X. It's an option not many think about, but it's there. If that is not feasible, make a new mechanic, but make it very clear and simple. Last thing you want is someone having to look for a table every time they want to say, turn undead.

Craft (Cheese)
2012-12-16, 09:24 AM
Unwarranted Statistical Advise:
If you're just using attack roles to determine success (and not damage bonuses or anything like that) there's no really good reason to worry about a bell curve. If your probability of success is 1/2, it doesn't matter whether your generating a normal random number and seeing if it's above the median, or flipping a coin and seeing if it's heads. The distribution of success and failure is exactly the same. Percentile dice make your probability of success a lot easier to calculate, and allow a lot more granularity than 2d6. Since the distribution is the same, why not use them?

When you're comparing static bonuses against a static DC they're equivalent, but when both the bonuses and the DC can change over the course of an encounter they aren't. A bell curve roll makes small differences count for more and big ones count for less. Which may or may not be what you want considering what type of game you're designing, but there's definitely a difference.

warty goblin
2012-12-16, 12:39 PM
But why use them? Percentile dice don't match my system and they generate big numbers. By using 2d6, I can make sure about 50% of a hit is luck and about 50% is skill (for an average person, against an average person) and that the difference luck makes goes down as you become more skilled.
Attack rolls in my system are made by rolling 2d6, adding two stats (they go from 1-6 each) and comparing it to the target's Defense score. I could change that to use percentile dice, but I would have to multiply scores or change the range of scores. Not very helpful.

Fair enough. Using a flat distribution does make it easier to fine-adjust probabilities of success though.


In my game, critical hits come up when you beat a target's Defense by a margin of 5 or more. So the number you roll matters. Not really, then you're just looking at single multinomial with three outcomes (miss, hit, crit) instead of a binomial (hit/miss).



Looks like a very small difference from where I'm standing, since the average of 2d4 is only higher than the average of 1d8 by 0.5, but I'm sure I'm overlooking something.

The difference in mean isn't particularly great. The variance reduction however is massive. 1d8 has a variance of 5.25. 2d4 has variance 2.5. If you continue buying expertise, the differences become substantial. 1d12 has variance 11.91. 3d4 is only up to 3.75. Makes something like damage rolls a lot more predictable, which has always been my complaint with skill-as-dice systems. More skilled characters become less predictable, not more. A master at something should routinely produce excellent quality outcomes, not have fairly high chances of getting 2's.

Even the d4 expertise buy doesn't really solve the problem though, and the increasing central tendency annoyingly means that slight increases in difficulty for 'average' tasks have massive impacts on the likelihood of success, while bumping a task from very hard to nearly impossible makes hardly any difference. You also run into oddities like a master chef being just as likely to cook a fantastic meal as they are of failing to open a can of soup.

What a person really needs for this is a way to skew distributions. Then a left-skewed highly skilled character is much more likely to do close to their best, while a right-skewed unskilled character is in the opposite position. On opposed rolls the skilled person will then demolish the amateur. For rolls against a fixed number, this still could be simulated with a flat distribution, but it's hard. You need to invert the cdf, apply it to the dice roll, and then compare that with the target number. Since the cdfs are likely to be ugly, this gets complicated.

Getting a skew with dice is tricky. Increasing your minimum is one way to get a right skew, so instead of 1d6 you roll Max(1d6, 2). Flat bonuses with a cap give you the left,so Min(1d6+3, 6). Both of these just put all the extra weight on the very bottom and very top though, while leaving the others equal. Probably sub-optimal.

Craft (Cheese)
2012-12-16, 01:03 PM
Getting a skew with dice is tricky. Increasing your minimum is one way to get a right skew, so instead of 1d6 you roll Max(1d6, 2). Flat bonuses with a cap give you the left,so Min(1d6+3, 6). Both of these just put all the extra weight on the very bottom and very top though, while leaving the others equal. Probably sub-optimal.

Or you could just (3+skill)d6b3. Negative numbers can be represented as adding the absolute value of the final penalty, then taking the lowest 3 instead of the highest. (So for rolling at -2, you'd roll 5d6w3).

{table=head]skill | mean | standard deviation | skewness
-5 | 5.6111950588705986 | 1.964849009310973 | 0.8113834421682956
-4 | 6.09724365569273 | 2.150493970796602 | 0.7185303952583563
-3 | 6.7262088477366255 | 2.3636451172570805 | 0.6085715395912492
-2 | 7.56983024691358 | 2.603467984691876 | 0.4716683244912087
-1 | 8.755401234567902 | 2.846844445311498 | 0.28350765297728486
+0 | 10.5 | 2.958039891549808 | 0
+1 | 12.244598765432098 | 2.846844445311503 | -0.28350765297728336
+2 | 13.43016975308642 | 2.6034679846918802 | -0.4716683244912048
+3 | 14.273791152263374 | 2.363645117257085 | -0.6085715395912453
+4 | 14.90275634430727 | 2.150493970796599 | -0.7185303952583609
+5 | 15.3888049411294 | 1.9648490093109765 | -0.81138344216829[/table]

I could make a table of comparative results for opposed rolls too if you like.