Thursday, December 29, 2016

How to Make Rulings

Dungeonmaster Plato believed that philosopher-DMs must be raised in special enclaves, away from any corrupting influence, if they are to learn how to properly DM.
So, the OSR says this thing all the time.

"Rulings, not rules."

It means that we'd rather have a small, tidy core of mechanics and then improvise all the uncommon rules, rather than have a giant encyclopedia of rules.

There are many reasons for this, and most of them are good ones.  (And I won't go into the reasons here.)

So if a DM wants to be good at rules, what does she have to do?

Well, she has to bust out the rulebook and memorize it.  Even the rules about grappling.  Even the rules about how long you can tread water.  This is what you have to do if you want to be a DM Who Is Good At Rules.

So if a DM wants to be good at rulings, what does she have to do?

. . .

Well, that's a tough one.  There's not a lot of guidance out there.  More ink has been spilled describing the overland speed of donkeys on a taiga (summer) than on how to make effective rulings.

Part of the problem is a conceptual one.  Rulings begin where rules run out.  They occupy the gaps in a rulebook, outside of where most writers spend their time thinking.  And it feels a litter counter-intuitive to put a chapter in a rulebook titled "How to Write Your Own Rulebook".  It's like printing a cookie recipe on the inside of an Oreo wrapper.

I have a tattoo based on this painting of Diogenes.
By Gerome.
And yet, making rulings is a very important skill.  It's what separates mediocre DMs from excellent ones.

So here's my attempt.

First question when designing a thing: what are the traits of a good thing?

1. Rulings Should be Fast

This is probably the most important one.  One of the biggest advantage of rulings over rules is that the DM just says some shit and the game keeps going.  No consulting the tome.  No arguing about rules.  (Or at least, keep those things to a minimum.)

2. Rulings Should Give Expected Probabilities

The in-game fiction needs to match the player's expectations.

Other people will restate this as: Rulings Need To Be Fair, but I think that's a less useful description.

If I'm DMing Hobbits & Hobbitholes and I realize that there is no rule for jumping over Farmer Maggot's dog, I'll have to make up my own ruling for jumping.  And whatever ruling I come up with had better give Legolas a better chance of success than Gimli, because everyone knows that Gimli can't jump.

If I make the jump a simple Strength check, I've failed, because Gimli is stronger than Legolas.  I could make it a Dex check, or a Strength check with a racial penalty to dwarves because of their stubby legs--the details don't matter as much as the resultant probabilities.

Relative probabilities are the most important (Legolas > Gimli), but absolute probabilities matter, too.

If I make the ruling, and the players do the math before rolling, and they realize that no one has more than an 7% chance of jumping over the dog, they'll (rightfully) protest.  It's five people jumping over one dog at the same time.  There's no way that dog should have such a high chance of stonewalling the entire party.

Rulings Should be Consistent

Consistent with other rules: getting hurled against a wall by a giant is pretty similar to taking a bad fall.  You'll need to come up with a good reason why one does lethal damage and the other does subdual damage.  The similarities are too big to ignore.

Consistent with other rulings: If you've been allowing players to coup-de-grace fallen enemies with ranged weapons, you should allow enemy archers to do the same thing to the players.  (You asshole.)

Next question: What are some tips that will help us achieve these goals?

Tip: Just Say Yes

"Okay, well you showed up with all the right preparations, and you can take your time when attempting it.  I'm going to rule that you just automatically succeed."

If the PCs have exactly the right approach, let them through.  If there's no rush for time, let them through.  If there is no penalty for failure (i.e. they can just re-attempt if they fail), let them through.

Seriously, I love this one.  Every DM should use it more often.

And if part of you bristles at letting the moment sneak past without the hand of entropy grazing it, tell them to roll a d20 and let them succeed on any number except for a 1.

Tip: Keep It Simple

Use established mechanics whenever possible.  Try not to invent them from scratch.

If there is a similar mechanic somewhere else, translate it.

Turn multiple rolls into a single roll.  Turn single rolls into static numbers.

Resist the urge to involve multiple parts of a character sheet.  "Well, it's a social check foremost, so I'll let them apply the Cha mod, but it's also trying to intimidate someone by crushing a skull, so I'll let them include their Str mod as well, but they should get a penalty for every steel item they have in their inventory, and. . ." That--that right there--is bad.

The ruling should touch as few parts of their character sheet as possible.  The most important thing is that it gives expected probabilities (i.e. Gimli should not be a better jumper than Legolas).

Tip: Learn How Probabilities Work

Your new best friend: anydice.com.  Read the Documentation.

The difference between d20+d6 and d20+3 is damn tiny.

The difference between 2d6-keep-highest, 1d6+1, and 1d8 is also damn tiny.  (At least in terms of averages, and if these are damage rolls, the details don't matter that much.)

Rolling a d20 with Advantage is usually damn close to a flat +4 bonus (plus or minus a point).

Whenever you roll one die, you have a flat probability curve.  Two dice gives a pyramid.  Three or more dice give a bell curve.  The point is, bell curves favor the stronger party, flat curves favor the underdog.

Usually, the players are attempting things that they are likely to succeed at.  This means that bell curves are their friend.  But when they are trying to do something really tough, like fight that higher-level demon dragon, the bell curve suddenly turns against them.

d20 and 3d6 both have an average of 10.5, but the curve is very difference.

d20+2 chance of rolling 10 or higher: 65%

3d6+2 chance or rolling 10 or higher: 84%

And that +19% chance is basically the equivalent of a +4 bonus.  Extreme example, but you get the point.

Opposed d20 rolls (e.g. d20+Str vs opponent's d20+Str) are also weird like that.  They give a big advantage to the stronger party.  Compared to a single roll (d20 + your Str - enemy's Str) which gives a smaller advantage to the stronger party.

(This is why I like single-roll mechanics; I like to be surprised with underdog victories.)

NOTE: I got a little bit crazy when I wrote this last part and you should probably just skip it.  Seriously, just pretend the post ends right now.  I can't delete it because I like it, but I also recognize that no one probably wants to read it.

Tip: Build and Test Complex Mechanics Before Implementing Them

Sometimes you want to anticipate a ruling, before the game even starts.  You are basically writing an ad-hoc rule.  Use the same tips as when making a ruling, but hey, you're not in a rush.  You can take your time when designing the rule.

My method: (1) Design a rule with the average party in mind in order to give them the desired probability of success, then (2) test it with other sample parties to see if it gives probabilities that you want.

This is a little bit like code testing.

A little while ago I wrote a dungeon that was pretty likely to feature a cave-in.  How long can a party survive while trapped in a room?  And how fast can they dig themselves out?

I'd advise you not to calculate room volume, look up oxygen consumption rates, infer oxygen consumption rates for halflings, research how fast miners are excavated in emergencies, etc.  Down that road lies madness.

I'd also advise against using general asphyxiation rules, since they aren't likely to serve your purpose. (And most systems are way too lenient with how long people can hold their breath anyway.)  There's no reason why you can't write a custom rule for this room.

I decided that I wanted the following features for my test party:

  • four PCs and no NPCs.
  • the PCs have 10s in all their stats.
  • I want this party to have a 50% chance of getting out alive.


We make up some rules:

  • Up to four people can dig at a time.
  • Everyone can breathe for 3 exploration turns before they need to make Con checks to stay conscious.
  • Each turn spent digging will yield 1d8 successes if they succeed on a Str check, and 1d4 successes if they fail on the same Str check.
  • Proper digging tools can upgrade a die by one or two sizes, depending.  Shovel = +1 die size.  Pickaxe = +2 die sizes.
  • After a certain number of successes, a hole is cleared and fresh air immediately fills the room.
  • How many successes are needed to clear a hole?

It's like a math problem!

For the first three turns, half of the party succeeds and half fails.  As people lose consciousness, fewer and few people contribute to digging.

  • First turn: 2d8+2d4 = 14 successes on average.
  • Second turn: 2d8+2d4 = 14 successes on average.
  • Third turn: 2d8+2d4 = 14 successes on average.
  • Fourth turn: 1d8+1d4 = 7 successes on average.  (Two PCs have passed out by now.)
  • Subsequent turns: asymptotic = 7 successes on average (Each turn is half as many successes as previously.)
So we tally them up and there's our answer.  The players need to get 56 successes before they clear a path to fresh air.

We can clean that up a bit.  "56 successes" becomes "60 cubic feet" or whatever.

How about if the party was stronger than average?  For example, what if they all had 12 Strength.  Well, if you do all the math, four PCs with 12 Strength will make an average of 59.2 successes before they all die.  That's interesting, because that's less than I would have thought.  It seems to indicate that the system is relatively tolerant of Strength imbalances.  So a strong party wouldn't have a huge advantage, and a weak party wouldn't have a huge disadvantage, which is more-or-less what I want.

True, we don't know the actual distributions, but the average is good enough for now.  I don't have all the fucking time in the world.

What if the party had 6 PCs instead of 4 PCs?  If we do the math (hint: it's the same as the original except that there are three active PCs on turn 4 instead of two) we can see that they would get an average of 70 successes before they all died.  This is significant, because it means that a large party is much better than a strong party. 

What if the party is small, and only has 3 diggers?  42 successes.  They're fucked.  

(For example, a lone delver trapped in the collapse would have virtually no chance to escape.  But perhaps this is as it should be.  Who would delve alone?)
What if the party has a pickaxe?  64 successes.  The prepared party is better than the strong party, but not better than the large party.

You may have noticed that this resembles the mechanics for combat, a little bit.  This is intentional.  Letting players get a sense of how much progress they've made towards a goal helps them understand how close they are to success/failure, while still leaving them time to change tactics if their first approach doesn't work.

For example, if you find out that you aren't killing the dragon fast enough to keep it from eating your companions, you pull out your vial of green slime and hope that you aren't making things worse when you throw it in the dragon's mouth.

In the cave-in example, the parties who notice that they are rolling very poorly and the air is getting stale will probably think of something stupid/ingenious in order to expedite their escape.  They'll do something risky, or they'll use up a precious resource in order to escape, but they'll probably escape.  Although I wrote the cave-in to have a 50% chance to kill a test party, I suspect that it would only TPK a tiny fraction of actual parties, just because players have so many resources at their disposal.

3 comments:

  1. I like to call this "Rulings Should Give Expected Probabilities," making rules predictable. My goal, most of the time, is for players to say something like "yeah, I figured it would be something like that" or "that's what I was thinking" when I make a ruling.

    ReplyDelete