Han Solo: We’re caught in a tractor beam. It’s pulling us in.
Luke: There’s gotta be something you can do.
Han Solo: There’s nothing I can do about it, kid. I’m full power. I’m gonna have to shut down. But they’re not going to get me without a fight.
Obi-Wan: You can’t win. But there are alternatives to fighting.

The group decides to make a Stealth check to hide from the Imperials. Han Solo makes a check for the group, with a bonus from the smuggling compartments on the Falcon. Stormtroopers search the ship, but their Perception check fails compared to the Stealth check. An Imperial lieutenant checks the ships logs, but the Stealth check rolls forward and his Computer Use check fails compared to it (so the logs show that the ship was abandoned shortly after takeoff).

Darth Vader issues an order for a scanning crew. Before they arrive, the PCs emerge from the smuggling compartments. They lose that bonus from their Stealth result, but the result itself continues to roll forward and allows them to ambush the scanning crew when it comes onboard. The PCs decide to take their uniforms (giving them a disguise bonus to their Stealth result).

At this point, however, there’s a point of uncertainty: The operators in the control booth try to contact TK-421 over the radio. The PCs decide to make a Bluff check to convince him that their radio is malfunctioning. The check is a success and it even convinces the operator to open the control booth.

Now there’s a brief interruption while Chewie and Han take out the control booth operators and then R2-D2 hacks the Imperial network to learn the location of the tractor beam.

Obi-Wan splits off from the rest of the party to deactivate the tractor beam. Han’s Stealth check continues to ride forward, but Obi-Wan is off by himself so the GM asks him to make a separate Stealth check (which will also ride forward).

Continuing his hack, R2-D2 discovers that Princess Leia is held in the detention block and is scheduled for execution. Luke convinces Han to help him rescue Leia, but they decide to increase their chance of reaching the detention block undetected by continuing their pretense of being stormtroopers with Chewbacca as a prisoner. Since this is the same disguise that they used earlier, the result of their Disguise check also rides forward and continues granting them a bonus to their Stealth result.

The result easily lets them move through the halls of the Death Star. The GM tries to add a complication by having someone board an elevator with them, but Han waves him off by suggesting that their prisoner is too dangerous.

The PCs arrive in the prison block. The GM makes a Perception check (or possibly two) for the guard on duty: Their Stealth check holds up, but their Disguise check isn’t good enough. He’s suspicious. This is another point of uncertainty, but this time they blow their Bluff check (“Prisoner transfer from cell block 1138.”) and he’s going to check their story.

Combat! Laser fire everywhere!

But that triggers a new point of uncertainty: They’ve set off alarms and someone is calling to see what the problem is. Han tries another Bluff check… and blows it again. The Stealth check stops rolling: “We’re going to have company!”

Luke opens the cell door. Princess Leia wakes up, makes a Perception check against the riding Disguise check, and succeeds. “Aren’t you a little short for a stormtrooper?”

Luke rips off his helmet. And, at that point, the Disguise check also stops rolling forward. (The PCs are changing their approach.)

R2-D2’s Hack check and Obi-Wan’s Stealth check, on the other hand, are still rolling along merrily.

The Art of Rulings: Let it Ride

Go to Part 1

Misapplying the technique of multi-step action resolution results in a problem I refer to as rolling to failure.

Perhaps the most common example of rolling to failure is the way a lot of GMs handle stealth: They’ll have a PC roll Stealth checks every single time they move or every single time they encounter a new opponent. If the PC has a 75% chance of success on a Stealth test, after just three checks their chance of success has dropped to 42%. After five checks, it’s dropped to 23%. And the first time they fail a check, of course, their attempt at stealth is over. (And the odds get even worse if the GM runs each of these checks by having all of the PCs roll Stealth, all the NPCs roll Perception, and if even a single Perception check is higher than a single Stealth check the gig is up.)

The consequence, of course, is that the players will simply stop trying to be stealthy. And, as a result, an entire set of tactics (and the really cool stuff that can flow from those tactics) vanishes in a puff of smoke.

“Why do my players always just charge in mindlessly?”

It might be because you’re not giving them any choice.

LET IT RIDE

The solution here is to Let It Ride.

When a PC is attempting a particular endeavor, make a single check and let the result of that check ride forward: If they’re trying to ascend a cliff, make a single Climb check to determine whether or not they get to the top (even if it would take them several rounds or minutes or hours to do so). If the intention is to “sneak through the enemy base”, then you make one Stealth check for the entire op and it determines how stealthy you are for the whole thing. (And the answer may be, “Not very stealthy, so someone is probably going to spot you.” And that’s OK.)

CONTINUING THE INTERACTION: A common mistake when letting it ride is to have the GM just summarize the entire attempt. For example, a player says they want to sneak into the orcs’ camp, so the GM calls for a Stealth check, sees a success, and says, “Okay, you manage to slide past the guards patrolling the border of the camp, make your way to the chieftain’s tent, and grab the tiara.”

That’s certainly an option, of course. But what’s great about letting it ride is that a player can still make meaningful choices while their result continues to roll forward. By leaving the player in the driver’s seat, the GM might discover that they want to knock out the sentries or eavesdrop at a tent to learn the orcs’ plans.

These choices could also impact the success or failure of the endeavor. For example, if they choose a path through the camp with higher security, it’s possible that their check won’t be high enough to handle it.

FAILURE POINT: Speaking of failure, how can the GM determine at what point during an attempt failure occurs if they’re using the let it ride approach?

Where the result of the check is being sequentially compared against a series of variable challenges (like a Stealth check being compared to various sentries and guards), this can be relatively easy. (You get spotted by whichever guard first exceeds your Stealth check.)

But what about something like that Climb check for scaling a cliff? On a failed check, how far up the cliff have they gotten before they failed the check?

(As with any check, of course, you still need to determine what failure means: Does it mean you fall? That you have to catch yourself before you fall? That you get stuck? That you realize you can’t reach the top for some reason and need to make your way back down?)

In general, there are three approaches: You can make a judgment call based on whatever result you think would be most interesting. You could determine it purely randomly. (For example, you could roll percentile dice to determine how far up the cliff they’d gotten or how much of their money they lost at the casino.) Or you could use the margin of failure on the check to calculate the result (or inform your judgment call).

POINT OF UNCERTAINTY: Another technique I’ve developed when letting it ride is the point of uncertainty. At the point of uncertainty, the original test result has been put at the risk of failure, but the character has the opportunity to save the result by taking some sort of action.

The point of uncertainty may be the result of the original check actually failing. For example, while sneaking into the orcs’ camp the PC encounters a scout whose Perception score is high enough to spot them. Can they knock him out (or take some other preventative action) before he raises the alarm? (If you wanted a mechanical cue, you might determine that a small margin of failure triggers a point of uncertainty and only a large margin of failure results in an immediate catastrophe.)

However, a point of uncertainty could also be the result of some tangential action coming in and interfering with your original check in an orthogonal fashion. For example, you’ve made a riding Social check while running a con job. You’re in the middle of chatting up the mark at a fancy party when you spot a former mark across the room. If they spot you, they’ll ruin your new cover. Can you make a Stealth check to avoid being seen?

What the point of uncertainty lets you do is to add complications while the overall resolution continues pushing forward. (Watch any heist movie ever made to see why that might be useful.)

Note that you’re never making a Stealth check to maintain your Stealth check or making a Social check to maintain your Social check. If you’re sneaking around and you’ve reached the point where the only way to succeed is to try sneaking again, that doesn’t mean that your original attempt is at risk. It means your original attempt has failed.

ENDING THE RIDE: With that being said, when does one attempt actually end and a new attempt starts? Failure is certainly one way. If the PCs fail their check to sneak into the orcs’ camp, fight the orcs who spotted them, and then attempt to reassert their stealthy approach, then that’s a new attempt and they should make a new check.

In the absence of failure, the first thing the GM should look for is a change in approach. If the PCs are sneaking along, decide to ambush some orcs who failed to spot them, and get into a big melee with them, that’s also the end of their stealthy approach and if they later decide to go back to sneaking they should probably make a new check.

(Note that these are not the same thing as an interrupted attempt. If you’re climbing a mountain and you fight some yetis halfway up, that doesn’t mean that your Climb check to reach the top of the mountain has failed or that the PCs have changed their approach. It’s just been interrupted by the yeti fight. The original Climb check should ride on through.)

More rarely, the GM can also look for significant changes in circumstance. This gets pretty fuzzy, though, because it’s really easy to fall into the trap of, “Every sentry they encounter is a change in circumstance! New roll!” Which is, of course, the exact opposite of what you want to do. But in some cases a seismic event will occur and things really will change so much that it’s basically a completely different task. For example, if the PCs are sneaking through the orcs’ camp and halfway through it a huge platoon of fire giants attacks the camp, the GM might decide that “middle of a battlefield filled with fire” is sufficiently different from “sleepy orc camp at night” as to make it a completely different situation requiring a new check.

(Another technique you can use to handle changing circumstances is to just modify the original check. For example, if a PC’s cyberdeck gets hit with a nasty virus while they’re hacking a mainframe, the GM can just apply a penalty to the original check result to reflect their debilitated equipment instead of calling for a whole new check.)

Next: Has the original goal been met? If the PC was attempting to seduce Baron von Isbury and the baron has now been seduced, then that ride has come to an end. (Insert bad sexual pun about the next ride beginning here.) If they were climbing a mountain and they’ve reached the top, then they’ve successfully mounted the obstacle. (Insert bad sexual pun about mounting here… Okay, I’ll stop.)

In a similar vein, however, you can also look at whether a significant landmark or accomplishment has been achieved. This can get fuzzy, too. (“Every time you get past a sentry, you’ve accomplished something! Make a new check!”) What you want to look for are the accomplishments that can’t be trivially negated as the result of a later failure. For example, instead of having a single Climb check for the entire mountain you might have a separate Climb check for achieving each base camp: On a failed check, instead of being forced to retreat to the bottom of the mountain they can instead return to the previous base camp (making it a significant accomplishment to achieve).

Similarly, after the PCs have penetrated the outer defenses of an enemy compound and planted their remote access device on the mainframe, you might have them make another Stealth check to get back out of the complex.

MULTI-STEP vs. LET IT RIDE

You may have noticed that at this point I have now advocated for breaking a single intention into a multiple steps AND for resolving broad swaths of time with a single check. Aren’t those contradictory?

Not really. These are different tools, and they can be used in different ways to achieve different effects. But how do you know which one to use?

Well, to a certain extent, this is another one of those areas which explain why this is the Art of Rulings and not the Flowchart of Rulings. But here are a couple things to keep in mind:

First, notice that if you’re properly applying multi-step action resolution, then each vector terminates at a meaningful consequence or meaningful choice. You’ll also notice that one of the ways that a riding check can come to an end is if the PCs choose a new approach or if they’ve reached some from of landmark (i.e., there’s been a consequence). In many cases, these mechanical approaches are actually two sides of the same coin.

Second, any place where you’re rolling the same skill check repeatedly and the consequence of any failure is to effectively wipe out all of the previous successes… You should probably be defaulting to letting it ride at least 99 times out of 100.

(The exception to this are endurance checks – i.e., situations where eventually failure is assured and it’s a question of how long a character can stave off that failure. For example, how long can you stay conscious in a vacuum? How long can you keep swimming in a terrible storm? How long can you survive exposure in the middle of a blizzard?)

Finally, you usually can’t go too far wrong by boiling this down to a simple question: Which one sounds like more fun?

Addendum: Let It Ride on the Death Star

Go to Part 9

Go to Part 1

Another resolution convention which GMs habitually fall into without really consciously thinking about it is the belief that every action needs to be resolved in a single check.

This is not universally true, of course. Combat is an obvious counterexample. And, indeed, it’s often intuitively understood that actions taken in the physical realm should be broken down into discrete steps: You can’t walk through a locked door until you figure out how to open it. If you want to shoot a rocket launcher from a sniper perch at the top of a cliff, you first have to climb the cliff.

These techniques, however, can be fruitfully applied in a much wider capacity. When we fail to recognize that, we end up robbing our gaming experiences of the depth they could possess. And, in some cases, we end up struggling with action resolutions which should be relatively straightforward to adjudicate.

MULTI-STEP ACTION RESOLUTION

Before we delve into how a multi-step resolution can be designed by the GM, let’s first consider a few ways in which multi-step resolutions can organically arise during play.

The first, and perhaps most straightforward method, is an incomplete declaration of method. To take one of our earlier examples, a player declares that they want to climb to the top of a cliff. That action is resolved. Then they declare that they want to shoot someone from the sniper perch at the top of the cliff. In this case, the player’s intention was always to take their shot from the sniper’s perch, but they broke that intention down into separate chunks. There is an instinctual understanding that X has to happen before Y.

Second, a multi-step resolution can also be the result of a partial success or failure: Your intention is to get to the other side of the chasm, but your Jump check was a partial failure and you ended up clinging to the ledge on the far side. Now you’ll have to take another action in order to complete the intention.

The take-away here is that resolutions which are potentially multi-step can be of a variable length and, in some cases, will even collapse down into a single step. (To invert the example, you might think that someone would need to leap across the crevasse, grab the edge, and then haul themselves up. But then they roll an extraordinary success and that multi-step resolution conflates down to a single leap that carries them all the way across.

Finally, there’s action economy. Multi-step action resolutions often crop up in combat, for example, because there’s a hard limit to how much you can accomplish on any given turn: “I will go here this round so that I can go over there next time.”

Looking at these organic examples, we see that they tend to break down into a pattern of X then Y then Z – i.e., X needs to happen before Y can happen. This is similar to the end of Star Wars: The Force Awakens: The landing party needs to successfully make a hyperspace approach to the planet before they can deactivate the shields, and they need to deactivate the shields before the Resistance can attack with their X-Wings.

But another option is for multiple requirements to be met before you can attempt a subsequent action – X, Y, and Z all need to be accomplished before you can do A. This is, roughly speaking, what happens at the end of Star Wars: The Phantom Menace: Amidala needs to capture the Viceroy, the Jedi need to defeat Darth Maul, and Anakin needs to destroy the droid control ship before the Trade Federation can be forced to negotiate a new treaty.

The other thing to notice is that this can become very fractal: It takes X, Y, and Z to accomplish your goal, but accomplishing X means doing A, B, and C first.

THE PATH OF MEANINGFUL CHOICES

Now, let’s turn our attention away from multi-step resolutions that emerge organically and instead look at how the GM can deliberately choose to resolve an action through multiple steps. When should you do it? Why are you doing it? How many steps do you break the action into?

One way of looking at this is what the Angry DM refers to as a visible benchmark: When you’ve completed one step of the multi-step resolution process, there should be a clear benchmark representing progress towards the goal. For example, imagine a generic scenario in which a PC is picking the lock on a door and the GM decides to resolve it as a sequence of three Lockpicking checks. After the first check has been completed, the GM says, “Do you want to continue picking the lock?”

That’s nonsense. Why is the GM asking that question? It’s meaningless. And you can imagine similarly meaningless interactions: “You’ve convinced him a little more.” “You’ve driven a couple more blocks towards your appointment, do you want to keep driving?” “You continue following the tracks.”

If we look back at our organic examples, we can see how they naturally include visible benchmarks: Once you’ve climbed the cliff, you’re no longer at the bottom of the cliff. After your leap across the chasm comes up short, you are now clinging to the opposite side (giving you different options than you had before). If you attack someone in combat the question could easily be, “Do you want to keep attacking him?” but if you succeed on the attack then you’ve dealt them damage and if your attack failed then you’ve afforded them an additional opportunity to attack you.

Coming back to our meaningless examples, we can also add visible benchmarks to them: For example, you might model picking the lock on a door as requiring three checks in order to determine how long it takes them to get the door open, measuring that against either a hard deadline (there’s a guard coming around the corner) or a fluid one (combat is raging around them and every extra round it takes them to open the door is another round their comrades have to hold off the orcs).

The vast majority of the time, these benchmarks should be visible to the characters, but there may be some instances – like the approaching guard – where the benchmark is meaningful in the game world without the characters (or possibly even the players) being aware of why.

In fact, we can probably generalize this concept of “visible benchmark” to “meaningful consequence”: Each step of a multi-step resolution should have a meaningful consequence. (And it should preferably be meaningful whether the resolution of that step is a success or a failure, for the same reason that this is the gatekeeper for single-step action resolution.) And this often means returning to our familiar friend, the meaningful choice.

In other words, if you look at the totality of an action resolution and you break it apart at each moment in which there is either a meaningful choice or a meaningful consequence… those end up being the individual steps of the multi-step resolution.

VECTORS

Another way of looking at multi-step action resolution is what Technoir refers to as vectors. To paraphrase from the game, you have a clear vector to your objective when:

• The player’s description of the action is feasible.
• There is a clear path for the action. There are no obstacles the character must surmount first.
• The objective is a logical consequence of the action described.

A vector is, in short, another name for a method, but imbued with the conceptual idea of a straight line: Look at where the vector is being aimed. If it can’t hit what it’s being aimed at (because there’s an obstacle in the way), then you’ll first need to identify a vector which will put you into a position where you can hit what you’re aiming for.

This process can be almost absurdly obvious when you apply the thinking to a physical objective. For example:

• You want to go into a room, but the door is shut.
• You want to open the door, but the door is locked.
• You want to pick the lock.

LOCK → DOOR → ROOM

But once you understand the basic concept of vectors, the same logic can be fruitfully applied to more abstract situations. They’re great for modeling social encounters, for example:

• You want to convince the rep from LVC (Lunar Venture Capitalists) to fund your zero-point energy generator, but you need to convince him it’s profitable.
• You want to convince him it’s profitable, but you need to convince him it works first.
• You want to use your scientific presentation to show him it works, but he’s busy and is brushing you off.
• You want to fast talk him into listening to you.

FAST TALK → PRESENTATION → PROFIT → FUNDING

And although these examples have their vectors drawn backwards from the goal (I want to take my shot at X, where can I see X from?), it can be equally useful to draw vectors from the opposite direction (particularly for GMs designing a multi-step resolution). Take a research test to discover information on the Serpent Crowns of Valusia for example:

• A web search doesn’t reveal much, but does tell you that there may be more information in H.L. Menckel’s Beneath the Waves: Arcane Archaeology of the Mediterranean, a rare volume.
• Additional research at the local college library indicates that the only known copy of the book was recently purchased at auction by Johnny Marcone.
• A networking test reveals that Marcone frequents the Velvet Room.
• A seduction roll gets you past the bouncers at the front door.

And so forth.

HOT-SWAPPING VECTORS

One of the major conceptual advantages of this approach is that you can easily hot-swap vectors: Instead of picking the lock, you can seduce the concierge to give you the key. Or pickpocket the master key off the bellhop. Or break the door down.

SEDUCE → DOOR → ROOM

PICKPOCKET → DOOR → ROOM

BREAK → DOOR → ROOM

Alternatively, instead of going through the door, you could climb through the window. Or break through the wall. Or teleport inside.

Of course, this also works great with non-physical vectors: You can get the LVC rep to talk to you by fast talking him. Impressing him with your past accomplishments. Seducing him. Using a display of your zero-point energy device to amaze him.

I find this concept of “hot-swapping” incredibly useful: It allows the GM to construct a framework for resolving complex, multi-step sequences without constraining the options of the players. It keeps your adjudication flexible and loose, allowing player creativity to flow through the structure.

Correctly interpreted, it also shows that the distinction between “organic” multi-step actions and GM determined multi-step actions is, in many ways, a purely arbitrary one. The “organic” examples are simply those where the GM and/or the players instinctively see the vectors involved, whereas the “determined” actions are simply those where they need to think about it. Over time, and with practice, more and more of these interactions are likely to become instinctual and second nature.

BREAKING VECTORS

In general, a vector should terminate at the point from which the next vector is being launched (i.e., the point at which the action changes direction). If you finish resolving a vector and the next vector is pointing in the exact same direction, you’re generally left with one of those meaningless questions we talked about earlier. (“Do you want to keep picking the lock?”)

Partial successes and failures, however, can often be expressed as broken vectors: You were running towards point X, but you slipped and you fell. Or something blindsided you. Or you smashed into an invisible wall (or other obstacle you were unaware of). In some cases, these broken vectors will create obstacles which will force the creation of new vectors to route around them, but in many cases they’ll be transitory delays after which the character can point themselves back in their original direction.

For example, let’s go back to that locked room:

LOCK → DOOR → ROOM

On a partial success, the character picks the lock to the door but is spotted by a security guard. This inserts a new vector, after which they resume their original trajectory:

LOCK → FAST TALK GUARD → DOOR → ROOM

Broken vectors can also be found in situations of endurance. For example, if a character is trying to hold a door shut while a werewolf pounds on it from the other side, they can end up with a vector that looks like this:

HOLD DOOR → HOLD DOOR → HOLD DOOR → HOLD DOOR

Probably repeating the same check over and over again from one round to the next while their friends desperately try to figure out an escape route.

You can see a similar pattern in what I refer to as operatic actions for purely idiosyncratic reasons (because I perceive a pattern in opera music where emotional crescendos are achieved through a series of cyclical builds in the power of the music). I also see this pattern a lot in anime or manga, where a character has to build up power over time and the longer they can sustain that build the more effective the result. (A more mundane example might be convincing members of the jury.)

However, now that we’ve talked about how to break an action resolution down into multiple parts, let’s do the exact opposite…

Go to Part 8

Go to Part 1

Many GMs have been conditioned to believe that there are two possible outcomes to every skill check: The character has an intention that they are attempting to achieve and they will either succeed at that intention or they will fail: You either walk the tightrope or you fall and you die.

In reality, when we make the decision to roll the dice, what we’re actually saying is, “There is more than one possible outcome to this action. Let’s find out which one happens together.” It’s a fictional cleromancy – a casting of lots to determine a fortune. And when we say that our cleromancy can only have two possible outcomes, we are limiting the efficacy of that fortune telling.

GRADUATED RESULTS

The easiest way to step away from the simple success/fail dynamic is to assign a single difficulty number, but then interpret the result based on the margin of success or the margin of failure. A very simple, universal metric of results looks like this:

Success
Partial Success
Partial Failure
Failure

In D&D we could set our margins to 5. If you succeed on a check by less than 5 points, you’ve scored a partial success. If you fail on a check but your result is within 5 points of the DC, you’ve only suffered a partial failure. The key idea behind any partial result is that it does not convey the full benefits of success or the full penalties of failure (and it will often carry with it the possibility of taking additional actions to improve your result).

For example, a character might attempt to leap across a chasm. The GM calls for a DC 15 Jump check. If the player rolls a 20 (a margin of success of 5), they easily leap across the chasm and land on the other side. If they roll a 16, on the other hand, they’ll only score a partial success and the GM might rule that they successfully leap across the chasm but fall prone on the other side. Meanwhile, a result of 12 (a margin of failure of 3) could result in them coming up just short, but managing to grab the ledge on the far side (giving the opportunity of pulling themselves up). Only by rolling 10 or less (a margin of failure of 5+) would the character fall helplessly into the chasm.

It’s also obviously quite easy to expand this spectrum. (Attack rolls in combat provide a simple example: A margin of success of 5 might award +2 damage, a margin of success of 10 might award +4 damage, and so forth.) In some cases the entire concept of “success” or “failure” will evaporate entirely – there is only the question of how well (or how poorly) the character did.

Another way of looking at graduated success is that, when there are multiple possible results, the GM models that by assigning multiple difficulties. I often use this technique with Gather Information checks, for example:

Note that both methods are really just different ways of looking at the same thing. We could just as easily write up that Gather Information chart as:

SIMPLE SUCCESS TEST

Now that we understand that mechanical resolutions can produce a spectrum of results, we can also take the next step of realizing that this spectrum does not need to be all-encompassing: The possible results for a given skill check do not have to range from “abject failure” to “outstanding success”.

A technique I find particularly valuable is referred to as a simple success test in Eclipse Phase: It’s an action where we accept that the character is going to succeed. The only question is how long it takes them or how good their success is.

If we consider our previous discussion of Take 1, we can see how simple success tests can mathematically emerge in many resolution mechanics: Success can be guaranteed no matter what you roll, so the only purpose of rolling would be to determine the quality of that success. (You can see the same thing on our Gather Information tables above: A result of 9 on the skill check would result in failure – they learn nothing about the so-called “Robert”. But if the character has a +9 modifier on their Gather Information skill, then success is guaranteed: They will definitely learn something about “Robert”. It’s just a question of how much they learn.)

In practice, of course, we don’t always have to perform a mechanical calculation to justify the simple success test. We can simply decide that, for example, the characters are professionals and this isn’t the sort of task for which there is any meaningful risk of failure for professionals. At this point, we can either default to yes and declare the action a success, or we can turn to our fictional cleromancy to discover the degree of success the character enjoys.

(We could also hypothetically talk about checks in which failure is guaranteed and the mechanical check merely determines the degree of that failure. I’m generally leery of such an approach because it feels as if it is most often abused in order to enforce railroads and the like. But it’s not impossible to consider a situation in which a character could deliberately choose a course of action for which they know that there is no possibility of success. The Battle of Thermopylae would be an epic example of that in practice.)

FAILING FORWARD

Another way of looking at the simple success test is the concept of failing forward.

In its most basic form, failing forward is largely indistinguishable from the simple success test: Mechanical failure is described as being a success-with-complications in the game world.

(The key distinction, if any exists at all, is that with a simple success test the GM is making a mechanical ruling that failure is impossible before the die is rolled. Failing forward, on the other hand, is an interpretation of a mechanical failure outcome after it has been generated. But, in practice, this is a fairly fuzzy line.)

For example, Lucas is attempting to pick the lock on the file room door and fails his skill check. The GM decides that Lucas still managed to get the door open… but it took too much time and now he’s been spotted by the night watchman. Or his lockpick broke. Or he’s gotten caught on camera and the bad guys will be able to track him down later.

Basically, there’s a large body of useful techniques you can explore as you break away from the basic success/failure paradigm. However, I would like to offer a few words of caution when it comes to the concept of “failing forward” because the terminology has attracted a few pernicious ideas.

First, failing forwards has become curiously fetishized by some players who believe that it should be used every single time. This seems to be primarily the result of people believing that failure automatically causes scenarios to grind to a halt. The classic example proffered is failing to find a clue and having a mystery scenario grind to a halt.

As the Three Clue Rule demonstrates, however, the solution to this problem is to offer multiple paths to success. And being forced to route around the roadblocks created by your failures will take you in directions you never anticipated: If you hadn’t failed to bribe the guards into letting you in through the back door, you never would have climbed the walls of the castle, broken in through the window, and fallen in love with the princess you found there. Failure is often the starting point of the most exciting situations and the most memorable of stories. Taking it entirely off the table won’t enrich your games; it will impoverish them. Like railroading, it’s a broken technique being applied as a hasty patch to another broken technique.

Speaking of railroading, the other major problem with “failing forward” is that it has accumulated a large amount of baggage from GMs who want to use it in order to keep the PCs on their railroad. (This may, in fact, be the origin of the term: “Forward” being the direction the pre-planned plot is supposed to be going.)

But neither of those problems are inherent to the basic concept, and the basic concept can be a very useful tool to tuck away in your toolkit.

Go to Part 7

Go to Part 1

The act of turning to the game mechanics is, ultimately, an assessment that there is variability in the potential outcome of an action. At the simplest level, we are saying that there is a chance the intention will succeed and a chance that it will fail.

Before we pick up the dice, however, we should take a moment to consider the potential failure state: Failure should be interesting, meaningful, or both. If it is neither, then you shouldn’t be rolling the dice. The clearest example of this is when the response to failure is to simply try it again:

Player: I try to pick the lock.
GM: You fail. What do you do?
Player: I try to pick the lock again.
GM: You fail. What do you do?
Player: I try to pick the lock again.

This is the gatekeeper of mechanical resolution. If the gate is locked (i.e., failure is neither interesting nor meaningful) then you should go back to the spectrum of GM fiat and remember to default to yes.

(It’s equally true that success should be interesting, meaningful, or both. But this generally takes care of itself because the players are not going to propose actions they are not interested in achieving.)

A common mistake GMs make, however, is to think that expending resources is automatically meaningful. For example, the most basic resource that one can expend is time. So they’ll look at the lockpicking example above and conclude that the failed checks are meaningful because they chew up time. However, this lost time only becomes truly meaningful it has consequences (i.e., wandering monsters, time ticking down towards a deadline, enemies on the other side of the door having more time to prepare, etc.).

The actual process by which an action check is made is obviously dependent on the game system you’re using. I’m not going to attempt a complete survey here, but what this usually boils down to is identifying the skill and setting a difficulty.

IDENTIFYING THE SKILL

Identifying which skill to use is pretty straightforward: Each skill will have a description which defines its parameters. You simply need to figure out which skill’s parameters the proposed action fits, and this is usually obvious.

In some cases, you’ll find that the proposed action can fall into the purview of multiple skills. Generally speaking, you can just let the character use whichever skill is better for them. The exception is if you feel that one of the skills is less related to the task at hand than the other: Systems vary in how they handle this, but allowing the check to be made with the alternative skill at a slight penalty is usually a good one-size-fits-all solution. (Another option is to allow a skill check using the alternative skill to grant a bonus to the primary skill. Or, as in D&D 3^rd Edition, allowing the character’s expertise in the secondary skill to simply provide a synergy bonus without any check.)

My personal preference is for systems that don’t have a lot of overlap in their skill descriptions. Some overlap is basically unavoidable, but being able to clearly call for a specific check generally streamlines the action resolution process by eliminating the back-and-forth of figuring out whether or not a particular skill would apply to this particular check. This is also why overlapping skills that are frequently used “in the blind” – like a Spot check to notice ambushers – are a particular pain in the ass: Since the player doesn’t know exactly what the check is being made for, they can’t let the GM know if they have an alternative skill they could be using: The GM calls for a Spot Tusked Animal check to notice the brain-eating walrus, but it turns out that the character actually has Spot Carnivorous Sea Mammals at a higher rating.

(Not an actual game. But it should be.)

Not all games have skills, of course. In most of those cases, however, you’ll generally follow the same basic procedure using attributes instead. (In many systems, skills and attributes are actually the exact same thing using different names: You take a single “this is how good I am at doing things” number and you want more detail, so you split it into a half dozen attributes. But then you still want more detail, so you split each attribute into a half dozen skills. It’s only when you get systems that freely pair skills with multiple attributes that the mechanic actually shifts. But I digress.)

SETTING DIFFICULTY

There are basically two ways of assigning difficulty:

1. Look at a list of difficulties and assign the difficulty by either description or analogy.
2. Start with a “default” difficulty and adjust it by considering the factors that modify that difficulty.

Some systems lend themselves more readily to one approach or the other. For example, D20 systems lend themselves to assigned difficulties and include difficulty tables that say things like, “A Hard task is DC 20.” or “A Formidable task is DC 25.” Call of Cthulhu, on the other hand, lends itself to adjusted difficulties by setting the default target number to the character’s skill rating so that the GM adjusts difficulty by applying a modifier to the rating.

Regardless of the system, however, you can use either technique. (And, in practice, you are likely to use combinations of both.) For example, when running D&D you could easily start with a default difficulty of DC 15 and then say, “Okay, it’s been raining and the rocks are slick, so let’s bump that up to DC 18.”

TAKE 1 / TAKE 10 / TAKE 20

When considering difficulty, there are three additional metrics I find useful. I’m going to use D&D 3^rd Edition terminology for them because that was the system where my thinking on this first crystallized. (Players of 4^th or 5^th Edition may find this confusing because the designers made a really weird decision regarding the handling of “passive” checks such that the description of the mechanics don’t match the mathematics of the mechanics. You’ll just have to suck it up, because I’m not going to try to jump through the broken hoops of poor mechanical design.)

TAKE 20: When you Take 20 in D&D, the result is calculated as if you had rolled a natural 20 on a d20. In other words, it’s the best possible success that the character is capable of achieving. It’s used in situations like our lockpicking example: The character is free to repeatedly attempt the task until they succeed, which means that we can just check the Take 20 to see if it’s a success or not.

TAKE 10: You can Take 10 in D&D when you’re not under any pressure. It’s the average result possible if you were rolling the dice, but the mechanic basically says “this is the level of success the character can achieve if they’re not under pressure or pushing themselves”.

TAKE 1: This concept is not labeled as such in D&D, but it flows naturally out of the mechanic. If you Take 1 on your roll, then it’s the worst result the character can have. If the difficulty of the task is equal to or less than the character’s Take 1, then the character will automatically succeed on that task.

Basically, these concepts break tasks down into three states: What characters succeed at without evening trying (Take 1). What they always succeed at if they make the effort (Take 10). And what they will eventually succeed at if given enough time (Take 20).

(For example, imagine that there’s something hidden in a room that requires a DC 25 Search check to find. A character with Search +5 will always find the item if they take the time to ransack the room. A character with Search +15 will find the item if they just quickly poke around the room. And a character with Search +25 will notice the item just by walking through the room.)

These concepts are generally useful in D&D (and other systems) for streamlining action resolution. But they can be specifically useful when setting difficulty by considering the type of person who would be attempting such actions and then using them as the analogy.

For example, I constructed these tables for D&D 3^rd Edition:

SKILLED PROFESSIONALS

GENERIC DIFFICULTY CLASS

TAKE 10 TRAINING: Ask yourself, “How much training would it take for someone to be able to succeed at this task as a matter of routine?” Find that level of training on the table and then add 10 to determine the DC of the check (as summarized on the Generic Difficulty Class table).

Example: Even someone without any training in pottery should be able to make a simple, crude bowl if they’re shown how the equipment works, so making such a bowl should only require a DC 10 check (0 + 10 = 10). On the other hand, it takes some training before someone should be able to reliably perform a backflip, so performing a backflip might take a DC 12 check (2 + 10 = 12).

TAKE 20 TRAINING: When dealing with particularly difficult tasks the question to ask is, “How much training would a person need in order to even have a chance to succeed at this task?” Find that level of training on the table and then add 20 to determine the DC of the check.

Example: An average person can’t just pick up a paperclip and pick an average lock. It takes training. So opening an average lock should be a DC 25 check (5 + 20 = 25).

Even if you’re not performing this mental calculation in the moment, this can still be a good exercise to familiarize yourself with what different difficulty numbers really mean in a new system. (I find these techniques particularly useful if you’re trying to calibrate difficulty ratings for characters outside of the human norm.)

But don’t use the character as their own analogy! Setting difficulty by looking at the stats of the character attempting the action and then calculating what you want the percentage of success to be is a pernicious practice. It can seem like a good idea because you’re gauging what an “appropriate” challenge would be for them, but the end result is to basically negate the entire point of having mechanics in the first place.

Some systems – like D&D or Numenera – lend themselves easily to this kind of analysis. Other systems, however, will obfuscate it. This is often true of dice pool systems. For example, the 2d20 System we use in the Infinity RPG uses a base dice pool of 2d20 which can be expanded through various mechanics up to a maximum pool of 5d20. The target number you’re trying to roll equal to or less than for a success is determined by the character’s skill rating, and the difficulty of the task is rated in the number of successes you need to roll: No matter how skilled you are, there’s no minimum level of guaranteed success. Nor, because of how the ancillary mechanics are designed, is there really a cap on the maximum success you could theoretically achieve.

You could still crank through a bunch of math and get some decent guidelines for dice pool systems like this, but in general you’re probably better off accepting the nature of the beast and using the adjust-from-default method of setting difficulty.

The 2d20 System largely sidesteps these issues, actually, because it doesn’t rely on the GM setting difficulty levels: At least 95% of the time the GM is basically deciding whether the task is of Average (1) difficulty or Challenging (2) difficulty. (Difficulty ratings of 3, 4, and 5 also exist, but are extremely rare in their application.) This is because the system is far less interested in the simple binary of passing or failing the check, and is instead intensely interested in the margin of success the character is achieving.

Which is exactly what we’re going to be discussing next.

Go to Part 6