How does an encounter begin?

It’s easy to fall into a simple formula: The encounter begins at line of sight (“you see an orc”) and immediately launches into an initiative check (“the orc tries to kill you”). There are usually minor variations on the line of sight (the orc opens the door, you open the door, the orc comes around the corner, etc.) and also the possibility of an ambush (you see the orc, but the orc doesn’t see you; or vice versa), but the formula remains pretty straightforward.

One way to break away from this formula is to vary the creature’s reaction to the encounter: Instead of leaping into combat, they might be friendly or attempt to negotiate or beg for help. In 5E Hexcrawl, I discuss how a mechanical reaction check can be used to prompt these disparate agendas.

I also talk about this a bit in the Art of Pacing, and also look at how shifting the bang – the moment at the beginning of a scene which forces the PCs to make one or more meaningful choices – can significantly shift the nature of the encounter:

Does the scene start when the ogre jumps out and snarls in their face? Or does it start when they’re still approaching its chamber and they can hear the crunching of bones? Or when they see a goblin strung up on a rack with its intestines hanging around its ankles… and then the deep thudding of heavy footsteps fills the corridor behind them as the ogre returns for its meal?

You can see how each of those creates a different encounter, and in most of them the scene starts before the ogre enters the PCs’ line of sight.

RANDOM ENCOUNTER DISTANCE

Another mechanical prompt that can help break the “line of sight” habit is a random encounter distance. This is a mechanic which dates back to the earliest days of D&D, but has faded away in more recent editions. But it can be a useful one in an game.

In the original 1974 edition of D&D, an encounter would begin at 2d4 x10 feet (or 1d3 x 10 feet if surprised). Many OSR retro-clones modify this to 2d6 x 10 feet.

This simple mechanic is largely all you need, neatly prompting you to think about how encounters begin in unusual and unexpected ways:

• If the distance generated is longer than line of sight, this suggests the encounter begins before the PCs can see the creatures (and vice versa), most likely because they can be heard (or their light seen around a corner).
• If the distance is closer than line of sight, what could explain the close proximity? (This is how you get moments like xenomorphs climbing through the ceiling panels.)
• In the case of wandering encounters, the result may also indicate the direction of approach: If one entrance to the room is 80 feet away and the other entrance is 20 feet away, and then you roll a random encounter distance of 20 feet… well, you can be pretty sure which entrance they’re using.

There may, of course, be times when common sense and the particular circumstances of the current situation will need to override the simplicity of this mechanic (but that’s why the GM exists in the first place).

CALCULATED ENCOUNTER DISTANCE

There’s another mechanical approach to this technique which was secretly hidden in 3^rd Edition D&D.

See, in 3^rd Edition there was a -1 penalty to perception-type tests per 10 feet. This meant that if you succeeded on a perception-type test, you could directly calculate the distance at which you detected the encounter by multiplying the margin of success by 10 feet.

For example, if an ogre rolled a Hide check of 15 and you rolled a Spot check of 24, then you’d have a margin of success of 9 and would detect the ogre at 90 feet. If, conversely, you rolled a Hide check of 18 and the ogre rolled a Spot check of 20, the ogre would be able to detect you at 20 feet. (Which, of course, means you would detect the ogre before the ogre detected you. The larger margin of success sets the encounter distance.)

You can follow this same basic guideline in 5^th Edition D&D and many other RPGs. And, once again, if the result is farther than the current line of sight, you’ll know that the opposition must have been detected in some other way (heard them talking, spotted in a reflective surface, etc.).

WILDERNESS ENCOUNTER DISTANCES

It should be noted that both the OD&D and 3^rd Edition mechanics don’t really work in the wilderness, which is why they included encounter distance tables for wilderness encounters, customizing encounter distances based on terrain type.

These tables were eliminated from the 5E core rulebooks (except, oddly, for underwater encounters), although they apparently appear on some of the official 5E Dungeon Master screens.

You can also find a continuation of these tables in Hexcrawl Tool: Spot Distances as seen here:

Thanks to the Alexandrites on my Discord and Twitch chat, who prompted and encouraged this tip.

Go to Part 1

THE FALLACY OF THE PERFECT CRIME

Something that can really trip you up when trying to create clues is the fallacy of the perfect crime.

For example, let’s say the PCs find an incriminating note that gives them a vital clue.

… but wait. Why would the conspiracy let the PCs find the note? Why wouldn’t they destroy it?!

Well, in some cases it’s carelessness. In others, they’re holding onto incriminating evidence that can give them leverage over their fellow conspirators. Sometimes they’re planning to destroy it, but they just haven’t yet. Sometimes the note has information they need for a job, so they hold onto them until the job is done. Maybe they have a sentimental attachment to the note because it was written by their dead wife. And so forth.

Alternatively, maybe they DID destroy the note and that’s the evidence that the PCs find. Sometimes this is structurally cosmetic (the only information to be gleaned is that the information has been destroyed). Sometimes this can be a consequence of the players’ actions, like in my Dragon Heist campaign where they knocked politely on someone’s door and then waited for them to open it while, inside, the bad guy was shuffling all their papers into the fire. (One of the nice things about the Three Clue Rule is that you can destroy clues with a clear conscience when it’s appropriate.) But often the PCs will be able to pull a clue out of the destruction (they burned the note, but this one enigmatic scrap remains!).

Of course, this logic also extends beyond incriminating notes.

Wouldn’t it make sense for the murderer to wear gloves or wipe away their fingerprints? (Maybe they did and the clue is something else. Or they tried, but missed a partial. Or they’re dumb. Or they panicked. Or they wanted to, but got interrupted.)

Wouldn’t they have disabled the cameras? (Maybe they didn’t spot the camera. Maybe they did hack the cameras and deleted the footage, but left traces from their hack that can be traced instead. And so forth.)

The perfect crime (or conspiracy), of course, would have flawless operational security and leave no evidence. But that’s not how it usually plays out in reality, and doesn’t really make for a great scenario in any case.

As GMs designing scenarios, I think we’re particularly susceptible to the fallacy of the perfect crime: Obviously if the bad guy knew they were leaving a clue, they would destroy it. You just thought of how the criminal could have left a clue, so obviously the criminal would have realized that, too!

What you actually want to do is the opposite: If you’re thinking about the crime and can’t find a clue that gives the PCs the information you need to give them… well, what extra mistake did the criminal make that will let you give them that information?

VARIED CLUES

As you’re designing clues for your scenario, you’ll want to make sure to include a wide variety of them. This is partly about creating a more engaging investigation. (If the PCs are just doing the same thing over and over and over again, that’s just less interesting than an adventure where they’re doing a lot of different things. And a puzzle isn’t really puzzling if the solution is always the same thing.) But it’s also structurally important: If all the clues are fundamentally similar, then it’s not just that the players are doing the same thing over and over again; it’s that the players MUST do that thing. And if they don’t think to do it, then they’ll miss ALL the clues.

The Three Clue Rule is built on redundancy, but clues which are overly similar to each other only provide a superficial redundancy. It’s kind of like monoclonal agriculture: When all the bananas are clones of each other, they’re all susceptible to the same pests and can be universally wiped out by a single disease. Just so with monoclonal clues, which can all fail simultaneously.

For example, if the PCs need to find out that Tony is hiding out at the Silver Rodeo, you might say:

• They can ask Tony’s girlfriend, Susan.
• They can ask Tony’s partner, Silvester.
• They can ask Tony’s mother, Sara.

Those are three different clues. But if the players, for whatever reason, don’t ask the people in Tony’s life where Tony is hiding out — because they don’t want to risk tipping him off, because they erroneously conclude that Tony wouldn’t have told anyone where he was going, or just because they don’t think of doing it — then that one failure will wipe out all of those clues.

Now, the principle of permissive clue-finding means that shouldn’t necessarily get rid of these “redundant” clues. But for structural purposes, they can be grouped together and only “count” as a single clue for the purposes of the Three Clue Rule.

When you’re looking to make varied clues, using clues with different forms is good. But the most important thing is that the clues should be found in different ways – different skills, different insights into how a crime scene should be investigated, and so forth.

As a final note, remember that the problem of monoclonal clues is limited to the clues pointing to a single revelation. It’s fine to design a scenario with lots and lots and lots of clues coming from talking to people, as long as those clues are spread out across a bunch of different revelations (each of which has varied clues of different types also pointing to them).

MULTI-PART CLUES

A common error that I see GMs make when playing around with the Three Clue Rule for the first time is to mistake the Three Clue Rule for a kind of logic puzzle:

• This clue indicates that the killer was wearing a green sweater.
• This clue indicates that the killer was taller than six feet.
• This clue indicates that the killer had gray hair.

If you combine those clues, you know that the only person with gray hair who was taller than six feet and also wearing a green sweater was Peter! So Peter killed Tony’s wife!

These clues can work if each uniquely points at Peter: He was the only one in the green sweater, the only one taller than six feet, and the only one with gray hair.

But if you need all three pieces of information – to eliminate the other people with green sweaters or gray hair or whatever – then this falls apart. Because you NEED all three pieces of information, that means that each piece of information (green sweater, six feet, gray hair) is actually a separate conclusion.

This doesn’t mean that you can’t design mysteries like this, though! Once you recognize that these are three separate conclusions, you can simply follow the Three Clue Rule: Have three clues pointing to the green sweater, three clues pointing to the killer’s height, and three clues indicating the killer had gray hair.

This technique can add a satisfying dimensionality to your mystery scenarios, giving the players a clear sense that their investigations are building towards some central revelation. It can be particularly effective for the Big Truth(s) in X-Files-type campaigns or identifying the location where the big conclusion of the campaign is going to take place. You can see an example of this in my Eternal Lies Remix.

FORMS OF CLUES

Let’s wrap things up by looking at the specific forms that clues can take. This won’t be a complete or encyclopedic coverage of the topic, largely because the cool thing about clues is that they can be virtually infinite in their form and variety, but hopefully I can provide a few ideas.

First, there are some broad categories of form that clues can fall into:

• Physical Artifacts
• Glyphs/data
• Bio-Signature
• Interrogation
• Surveillance

These, too, are not comprehensive. And the boundaries between them aren’t exactly razor-sharp.

Here’s a list of specific examples drawn from scenarios I’ve designed:

• Correspondence (letters, e-mail, etc.)
• Diaries
• Ephemera from a location (matchbooks, theatrical posters, tickets, etc.)
• Official reports
• Tracks
• Surveillance (of a person or location)
• Tailing someone
• Business cards
• Fingerprints
• DNA
• Blood type (including fantastical types like “Vulcan”)
• Graffiti
• Financial records
• Tattoos
• Canvassing
• Video/audio recordings
• Mystic visions/strange dreams
• Shipping information (tracking, postmarks, return addresses, etc.)
• Books (including inscription and marginalia)
• Bureaucratic records/background checks

In the Three Clue Rule, I assert that for each conclusion you want the PCs to make in a mystery scenario, you should include at least three clues. Many people have had a lot of success designing and running scenarios with this advice, but as GMs begin working with this technique for the first time, it can often prompt the question:

How do you come up with all those clues?

So let’s talk a little bit about what goes into making clues and a few of things I’ve learned over the years that you may find useful.

First, check out Random GM Tip: Using Revelation Lists. That article provides an in-depth look at how I organize my clue lists while designing and running scenarios, and that structure is going to be useful as you start laying the clues into your scenario.

Second, don’t overthink it. A lot of clues will flow naturally from the scenario as you’re designing it. I recommend doing an initial “structural pass” in which you’re defining the necessary revelations and assign clues to them. If you can get three clues for every revelation at this point, that’s fantastic. But if you’re struggling to fill in all the blanks on your revelation list, that’s just fine: Start working out the details of the scenario with your revelation list nearby and you’ll often find clues easily along the way. (“Oh! Silvester would definitely know that Uncle Bob used to be in the marine corps!” or “Hey! I could put a matchbook from the Silver Rodeo in Susan’s purse!”) And if that still doesn’t fill in all the clues you need, all this detailed knowledge of the scenario will make it a lot easier to come up with the clues you still need during a second brainstorming session.

SIX TYPES OF CLUES

Conceptually, there are six types of clues.

Static clues are a specific piece of information which can be found in a specific way. For example, a bloodstain that can be found by searching a room.

Flexible clues are pieces of information that you know are there to be found, but which can be accessed in multiple ways. For example, imagine that there’s valuable information in a computer database. The PCs could hack the database remotely, physically raid the building where the server is located; bribe, seduce, con, or otherwise subvert an employee with access to the database; and so forth.

Broadly speaking, PCs will learn about people, places, organizations, events, or other scenes where they can investigate whatever it is that they’re investigating. In the terminology of node-based scenario design, these are nodes, and static and flexible clues are placed within specific nodes.

Proactive clues, by contrast, come looking for the PCs. I talk about them at some length in the Three Clue Rule and also The Secret Life of Nodes, but Raymond Chandler provides the classic example: “A guy with a gun walks through the door.” In other words, instead of the PCs identifying a node and going to investigate it, some nodes (usually people) will come to the PCs and bring clues with them.

Reactive clues, on the other hand, kind of exist “in the cloud,” so to speak. They aren’t assigned to specific nodes and they don’t come to the PCs. Instead, these are clues the PCs will find through holistic investigation techniques that they can frequently use without specific prompting. Typical forms include canvassing, research, and divination spells. Rulings in Practice: Gather Information dives into these types of clues in greater depth.

(This terminology can be a little weird. You can think of it like this: If the clues are proactive, then the PCs are reactive – i.e., the clue walks through the door with a gun and the PCs have to react to that. If the clues are reactive, then the PCs have to be proactive – nothing is going to specifically prompt them to go looking for these clues.)

Because reactive clues are more-or-less totally dependent on the players spontaneously thinking to go looking for them, they can be very unreliable when designing an adventure. On the other hand, if looking for a particular type of reactive clue becomes a standard operating procedure for a particular group, then the exact opposite can be true!

For example, over the past couple of years I’ve spent a lot of time running the same Call of Cthulhu and Trail of Cthulhu scenarios for different groups. Groups with players who have read the rulebooks or have a lot of experience with the game will almost automatically go to the library or research the local newspaper morgues for references to whatever they’re investigating because the rulebooks establish this as a standard operating procedure for investigators. Groups without that experience just… don’t.

The Technoir roleplaying game is actually mechanically designed around the default action of “hit up one of your contacts.” If the PCs don’t know what to do next, they should go ask one of their contacts (and the game is designed so that the contact always has a clue or a job or some form of lead that gives them something to pursue). The first few times I ran the game, this worked flawlessly. Then I ran the game for a group that hadn’t read the rulebook and didn’t know to do this and the game turned into a grind for a couple of sessions until I realized that I just needed to literally tell new the players: If you’re stuck, do the noir thing and hit up a contact!

In other cases, you’ll find groups spontaneously developing their own standard operating procedures.

Dynamic clues are what the PCs find when they take an investigative action which should logically provide information despite the fact you didn’t specifically prepare a clue for it. This is what the Three Clue Rule refers to as permissive clue-finding. For example, the PCs decide to look around outside the house. You didn’t specifically anticipate them doing that, but you know that the killer ran into the tree line at the far end of the property, so you conclude that the PCs could potentially find the killer’s footprints.

However, not all dynamic clues are the result of you being blindsided during a session. In some cases, they can actually be designed into the structure of a complicated scenario where you can be fairly certain the PCs will take a particular class of action, but you can’t really sure exactly what form that action will take. For an example of this, check out the Dragon Heist Remix, where, for example, the PCs might choose to research a particular faction:

If the PCs want to find a faction by doing general research, point them in the direction of one of the faction’s outposts. (Each outpost will contain clues that point to Lairs, which are generally their ultimate goal.)

And similar guidance is given for what happens when they track or interrogate bad guys. (You could also think of dynamic clues as being flexible clues that are flexible to the point of formlessness.)

Unassigned clues are basically what happens when you go whole-hog on permissive clue-finding: Instead of prepping specific clues, the GM only preps a revelation list. In some cases these revelations will be assigned to specific scenes (i.e., the fact Tony is hiding at a lake house can be found at the Silver Rodeo). In either case, the GM waits for the PCs to propose any investigation action, then looks at their revelation list and improvises an appropriate clue for that revelation.

I tend to be fairly skeptical of this approach:

• It puts A LOT of pressure on the GM’s ability to improvise during sessions.
• It often produces wishy-washy scenarios, that tend to lack texture. (A good mystery scenario is often as much about what you don’t find as what you do, and this technique tends to miss those beats in the story.)
• As mentioned above, many clues will tend to emerge naturally from the details of the scenario as you design it. So this technique usually goes hand-in-hand with underdeveloped scenarios.

I’ve run mystery scenarios that were entirely improvised, so the technique certainly can work. But I think you will almost always get better results by having a robust, reliable foundation and then improvising on top of it.

MAKING CLUES: STRUCTURE

When making clues, the first thing I do is think about what the PCs need to know structurally for the adventure to work. This is why I start with a revelation list: There may be a lot of other information that the PCs discover in the course of their investigation (that the villain beats her husband; the particular effects of the poison the killer is using; the cult’s beliefs on the wisdom of cats), but I really want to keep my focus on that essential structure.

Then I look at what I know about the crime – or whatever it is that they’re investigating – and think very specifically about what in the specific node I’m looking at could indicate the thing they need to know.

This may seem obvious, but I can get lost a surprising amount of the time floating around in a, “What clues are there?” haze. You really want to flip that around: Instead of thinking about what clues might be in this scene, focus on what you need the current scene to tell the PCs and then treat that as a puzzle or a problem to solve.

For example, the PCs are investigating a cabin on the lake. If you start by saying, “What clues would be in the cabin?” that’s too broad. It’s too vague.

But if you instead say, “I need a clue that points the PCs to Cai Lijuan,” that’s far more actionable:

• There’s a crumpled up envelope in the wastebasket addressed to Cai.
• The property owner can identify Cai as the person who was renting the cabin.
• There’s a box of cold pizza in the fridge. Cai’s name is printed on the delivery label.
• The family vacationing in the next cabin down the lake met Cai and knows his name.
• Cai’s car is still parked at the cabin; you can run the license plate or VIN numbers.

And so forth.

MAKING CLUES: THE SKILL LIST

For inspiration, look at the skill list in the game you’re using. Pick any skill. How could the PCs use that skill to get the information they need?

For example:

• Cryptography? Cai’s left his diary, which he writes in a personal code, in the bedside table.
• Locksmith? There must be something Cai locked up here. Let’s say there’s a safe hidden behind a picture frame with documents identifying Cai inside.
• Photography? There’s a USB stick with photos. Nothing identifying in the photos themselves, but you could check the EXIF data.

Obviously not every skill will be relevant to every piece of information. But if you have a particular piece of information in mind, running down the skill list will probably make specific skills jump out and provide ideas.

Go to Part 2

We’ve reached the end of combat and the last bad guy is down to the dregs of his hit points. One of the players makes an attack roll… He hits!

And the Game Master says, “Don’t bother rolling damage! He’s dead!”

Don’t do this.

Before we delve into why this is a bad idea, let’s first talk about why the impulse exists: The combat is clearly coming to a conclusion and the remaining combatant poses no meaningful threat to the PCs, so there’s no longer any tension or meaningful stakes in the scene; it’s been reduced to a rote resolution. Heck! The bad guy might only have one or two hit points left, so the outcome really IS predetermined here, so why bother rolling the damage dice?

This is pointless! Let’s wrap it up!

This impulse is not necessarily wrong. It’s just mistimed.

The key thing here is the ownership of the win. When a player rolls a successful attack, deals damage, and the bad guy dies, that’s something that THEY did. They own that moment.

If you, as the GM, interrupt that process, and declare a fiat success, you take that moment away from them: They didn’t kill the monster; you did.

It’s a subtle distinction, and it won’t always result in the moment getting deflated, but it’ll happen often enough that it’s worth steering clear of this technique. Particularly since the benefit you’re getting is so minor: You’re saving… what? Fifteen seconds by having them skip the damage roll?

OTHER OPTIONS

The obvious alternative it to just let them roll the damage and then announce the result.

That works if the damage kills the bad guy anyways. But what if the attack doesn’t quite deal enough damage to finish the job? This fight is boring! It’s time to be done with it!

First, double-check to make sure that’s actually true. As the GM, the fight has become boring because you can see the numbers and the outcome has become certain. In your role of playing the bad guys, you have lost meaningful agency and that’s boring. But your experience here may not mirror the players: they haven’t lost agency. In fact, they’re about to reap the rewards of their agency! They’re going to win! Winning is exciting!

Second, if it IS time for the fight to be done, the next easiest option is to fudge the bad guy’s hit points. I’m generally not a big fan of fudging, but it’s probably just fine here. You aren’t actually changing the ultimate outcome here (which is where all the various problems with fudging come from); you’re just speeding it up.

Another option is to not initiate the attack roll. When the player says, “I attack the monster with my sword!” you immediately assume they have successfully done that and describe the outcome. This is, in fact, in keeping with the Art of Rulings: The player has announced an intention to kill the monster. You know that this will definitely succeed. So the appropriate ruling is actually default to yes.

You might not think this would make a difference. It seems virtually indistinguishable from interrupting the damage roll! And there is still some risk here (from the declaration of intention, the player has a mental momentum reinforced by the rhythms of the combat system that’s driving them towards the attack roll, and interrupting that momentum can be disruptive), but in my experience it’s much less likely to cause a problem.

And you can enhance this technique by empowering the player’s agency: When they say, “I attack the monster with my sword!” you can ask the players to describe the coup de grace. “Agnarr’s mighty blow finishes off the goblin! What does that look like?”

Another alternative, if the combat is lagging and you’re concerned the current PC’s attack may not deal enough damage to end it if they roll poorly on the damage dice, is to tell them how many hit points the monster has left before they roll the damage dice. This is almost the exact opposite of fudging the monster’s remaining hit points, effectively blocking you from using that technique.

The reason this works is that knowing the damage needed puts the table’s intense focus on the damage roll: Everyone knows exactly what needs to be rolled, the tension will build as the dice are picked up, and then explosively release (either in triumph or failure) as the result is revealed. If the bad guy dies: Great! You’ve injected that moment with a little extra oomph!

If the bad guy doesn’t die? That’s okay! You’ve still clearly framed that this is the Final Countdown (so to speak) and that focus will tend to carry forward to the next player’s attack.

Yet another option is to remove the moment entirely by training your players to roll attack and damage simultaneously (as described in Fistfuls of Dice).

END IT EARLIER

Taking a step back, it can also be useful to consider how we got to this specific moment: How could we have avoided getting to the point where a combat encounter is ending on a whimper?

The obvious answer here is to end the combat sooner.

One option is to have the bad guys run away! The PCs take out the Big Boss and that’s the sign for the all the mooks to hightail it! Or, alternatively, the PCs take out a bunch of mooks and the Bandit King decides discretion is the better part of valor.

(Check out The Principles of RPG Villainy for a lengthier discussion of how and why having your bad guys run away is a good idea in any case.)

Surrender is another option, although that can have its own issues.

Alternatively, pursue the Default to Yes solution more aggressively: You don’t have to get down to the very last bad guy’s very last hit points to recognize that the encounter has reached its conclusion (and the end of the fight is now foreordained)! At that moment, a nice, clean option is to ask each player to describe what they do finish things up. (You can do this in initiative order to provide a little structure if that’s useful.)

A key thing here, though, is to make sure that there has actually been a conclusion! And that the players can feel ownership of that conclusion! (The example of the PCs taking down the Bandit King and then the mooks panicking is a good example: The combat ending is a clear result of the PCs doing something decisive and significant.)

I discuss this a bit in GM Don’t List #5, but a GM prematurely ending combats because they routinely get bored with the fights they set up creates a number of other problems:

• Characters built to enjoy their spotlight time during combat are being punished.
• Strategically clever players often spend the first few rounds of combat setting up an advantageous situation, and it’s frustrating if that gets prematurely negated.
• If players feel that encounters are being summarily dismissed (in a way that isn’t affected by their agency), their uncertainty about which encounters are actually being determined by their actions will make it difficult for them to determine when and how to spend limited resources. (Burning a one-use potion or once-per-day ability only to have its use become irrelevant is incredibly frustrating. If the use of that ability is what effectively ends the threat of the combat, make sure you emphasize that in framing the end of the scene so that the player’s agency is given its due.)

And, as I mentioned in that earlier essay, “All of these problems only get worse when the GM defines ‘boring’ as ‘the PCs are winning,’ while remaining fully engaged and excited as long as his bad guys have the upper hand.”

With those words of caution in mind, though, the art of knowing when a combat encounter is effectively done (or, perhaps, when a combat encounter is done being effective) is a really important part of your skill set as a GM.

(For a wider discussion of how to effectively end scenes, see The Art of Pacing.)

The key thing is to make sure that the players feel ownership over what happens: That THEY were the ones who won the fight.

If in doubt, have the bad guys run away and let it play out. Player agency can persist through that decision, so you have a much wider margin of error.

In “The Joy of Extradimensional Spaces,” an adventure by Michael Polkinhorn in Candlekeep Mysteries, the PCs discover a portal to an extradimensional space, go through it, and become trapped on the other side.

This is a classic trope, and for good reason. Perhaps the most memorable and well-known version are the arches of mysterious mists and the green devil faces of The Tomb of Horrors, but I’ve run some variation of this gag countless times.

Often, of course, PCs will also encounter mystic portals and strange gateways of coruscating energy that aren’t one-way affairs. But they won’t necessarily know that. And even if they don’t suspect a trap, the unknown threshold beyond which they cannot see is quite likely to inspire endless amounts of paranoia.

So, either way, the day will likely come when you hear some variation of: “Okay, let’s send Kittisoth through first to check things out.” Or: “I stick my ten-foot-pole into the portal and pull it back out.”

First: Know your metaphysic.

What does happen when you stick something halfway into a magical portal and then pull it out?

A few variations:

• No problem. It’s just like a doorway in space, even if the field of energy blocks your line of sight. (Perhaps you might even be able to still hear what’s on the other side.)
• You can’t pull it out. Once an object has gone partially through the portal, the only movement allowed is forward through the portal. If you pull back, it will feel as if the object is “stuck.”
• That’s incredibly dangerous. You can pull back, but you’ll only pull back the portion that’s still on THIS side of the portal. (Make sure you walk through these portals with confidence.)
• That’s not possible because as soon as any discrete object or creature touches the surface of the portal, it instantly vanishes and reappears on the other side. It’s more of a “touch here to activate” effect than it is a literal gateway. (This one can get tricky: If I prod it with a pole, do I vanish? Can I tie a rope to something hundreds of feet away, toss one end of the rope in, and have the whole thing go through the portal? What if the object is bolted to the floor? Or is a huge tree and the room on the other side is only 10’ x 10’? But, conversely, if I touch it with my hand do I disappear while leaving my clothes behind? One way to simplify this is to create a “threshold” before the portal – a misty arch, a field of energy, a penumbral aura, the entire room that the portal is in – and only objects within that threshold vanish when the portal is touched. Your choice whether to have something straddling that threshold get severed at the threshold or if the whole object remains even when touching the portal. But I digress.)

You want to have a clear understanding of this, because the players will want to experiment with the portal to figure out how it works / whether or not it’s safe. They’re going to come up with all kinds of crazy testing schemes, and you’re going to want a clear conceptual framework for consistently ruling what the outcomes of those tests will be.

I’m of the opinion that this metaphysic should feel consistent with the portal’s behavior on the other side: If it’s a one-way portal, for example, then you shouldn’t be able to stick things through the portal and then pull them back. (In other words, you want to play fair and reward players who take the effort to engage deeply with the scenario.)

Second: Use a portal countdown.

When the PCs send a scout through the portal, don’t immediately describe what the scout sees. Instead, stay with the PCs that remained behind: Describe the scout going through the portal, disappearing, and then… what do you do?

Keep track of the number of rounds it takes for each of the other characters to go through. Then, once they’ve all done so, you can flip to the other side of the portal and accurately play out events on the other side.

So, for example, Kittisoth walks into the portal. The other PCs wait a couple of rounds and then Edana sticks her ten-foot-pole through. Unable to pull it out, she sighs and steps through on the next round. Everyone else then nervously follows on the fifth round.

Now cut to the other side of the portal: Kittisoth emerges into a goblin ambush! Roll initiative! On the third round of combat, Edana’s pole sticks out of the portal. On the fourth round, Edana walks out. On the fifth round, the other PCs join the fray!

Even if there are no immediately “interesting” consequences from using the portal countdown (e.g., the portal emerges into an empty room and Kittisoth just patiently waits for everyone to join her), the experience is still immersive and tantalizing in its paranoia: What is happening on the other side of the portal?

When there IS something interesting happening on the other side of the portal, the experience of seeing the other side of the timeline (e.g., Edana’s pole sticking out of the portal) is a ton of fun.

What if all the PCs don’t come through? That’s fine. Run a couple rounds of the portal countdown in “combat time,” but when it becomes clear that some or all of the remaining PCs aren’t going to follow Kittisoth blindly through the portal (“She was supposed to come back!”) simply ask them how long they’d wait, note that on the countdown, and then cut to the other side of the portal.

Should I have players leave the room? If you’d like. I’ve not generally found it to be necessary and, as I mentioned, I’ve found the audience stance of seeing the other side of the timeline to be fun. But I’ve also been given to understand that the experience of going into another room and waiting while your fellow players join you one by one can also be mysterious and fun. Play it by ear and get a feel for what works.

(If you’re splitting the players up anyway, you may also want to just go back and forth between the two rooms each round instead of using a portal countdown.)

What if it’s not a one-way portal? This technique works best with one-way portals because the events on one side of the portal are generally firewalled from the other side of the portal, but since the players don’t always know if the portal is one-way or not, it can also be effective to use the technique regardless. (If nothing else, it means you’re not tipping your hand when they do encounter a one-way portal.)

But what if, for example, Kittisoth simply walks back through the portal?

If that outcome is quite likely, of course, then you don’t want to use the portal countdown technique in the first place. Or use it in a modified form. For example, if Kittisoth’s intention is to walk through the portal and then immediately turn around and walk back and there’s nothing on the other side that would prevent that from happening, you can just run it normally. Or you might take Kittisoth into another room, describe what she sees, and get a sense of her intentions: If it looks like she’s going to explore a bit – or get cut off from the portal by the goblin ambush – then you can go back and run the portal countdown.

But if you are using a portal countdown and Kittisoth comes back “early” (i.e., before all of the declared actions on the countdown have played out), that’s just fine. You can simply retcon the rest of the countdown; those actions belong to an alternative version of reality, I guess.

(The portal countdown is basically an extended form of declaring an intention. It’s similar to combat systems in which everyone declares their actions at the beginning of the round before resolving them in initiative order: If someone declares that they’re going to run down an open passageway, but then the passageway gets blocked with a wall of stone before they can take their turn, they’re not going to just mindlessly grind their face into the wall like a computer game NPC with bad pathing. They just won’t take the action.)

What about other continuity issues? Keep in mind that premature portal returns aren’t the only contradiction of continuity you can run into. For example, those on the far side of the portal might use a telepathic ability to contact their comrades. Or Kittisoth might grab the end of Edana’s pole and yank her through the portal.

Again, that’s fine. Just end the countdown and begin resolving actions normally from that point. (Which might even include starting a new portal countdown depending on what’s happening.)

Similarly, you generally pause the countdown and begin resolving actions on the far side of the portal when everyone has either gone through or definitively decided NOT to go through. But actions on the near side of the portal can also trigger this decision earlier: For example, if it’s Edana who chooses to use a telepathic ability to contact Kittisoth (assuming that she can do so), you’ll need to pause the countdown and resolve Kittisoth’s side of the portal.