CHAPTER 17: ENCOUNTERS FOR THE GM
An encounter is any moment that requires the party to make a decision.
That definition includes combat, obviously. It also includes the sound of footsteps in the corridor ahead, a door that opens onto a 6-metre drop, a merchant who stops the party in a market square with very specific questions, and a garrison patrol that hasn't spotted anyone yet. The fight is one resolution of an encounter, not the encounter itself. GMs who design only for combat produce sessions where the only interesting question is which attack roll lands first. GMs who design for decisions produce sessions where the party is navigating a world.
This chapter covers how to build those situations — for dungeons, for wilderness, for the particular texture of the Known World — and how to use the game's tools to make the result feel lived-in rather than constructed.
WHAT ENCOUNTERS ARE FOR
Every encounter serves one or more functions. Knowing which function a given encounter is meant to serve helps the GM design it well and run it honestly.
Resource encounters drain the party: HP, spells, equipment, and time. Their function is to make the things the party reaches for expensive, so that arriving at the dungeon's deep level with three spells remaining and one cleric at half HP carries real weight. A resource encounter doesn't need to pose a genuine defeat condition — it needs to cost something.
Decision encounters present a choice with meaningful consequences. The fork in the passage where the left corridor smells of smoke and the right has fresh drag marks. The NPC who offers the party a shortcut in exchange for something they'll regret later. The creature that's clearly dangerous but clearly not hostile, which the party can provoke, negotiate with, or avoid. Decision encounters reveal character and shape how the rest of the session unfolds.
Information encounters give the party something they need: a map fragment, a rumour that reframes what they thought they understood, a glimpse of what they're walking toward. These encounters often look like nothing — a dying bandit, a fresco on a wall, a creature that runs from the party — and GMs underestimate how important they are. A party that walks into the final confrontation completely blind is facing unfair conditions, not a difficult challenge. Information encounters are what make the challenge fair.
Atmosphere encounters build the world. The sound of something large moving in the water below the bridge. The tracks of an animal that the party can identify as being dead for several centuries. A shrine to a god the party's cleric recognizes, maintained recently despite the dungeon's apparent abandonment. These encounters don't advance the plot or drain resources — they tell the party what kind of place they're in.
Most encounters serve more than one function. A wandering ghoul is primarily a resource encounter, but it's also an atmosphere if the party finds it still wearing the uniform of a Second Era soldier. A merchant is primarily information, but becomes a decision encounter if the information she's selling is something one faction is very interested in keeping quiet.
ENCOUNTER DIFFICULTY
Sonata uses creature level — derived from XP value — to calibrate whether an encounter is appropriate for a given party. The table:
| XP Value | Creature Level |
|---|---|
| 1–20 | 1 |
| 21–50 | 2 |
| 51–150 | 3 |
| 151–250 | 4 |
| 251–500 | 5 |
| 501–1,000 | 6 |
| 1,001–3,000 | 7 |
| 3,001–5,500 | 8 |
| 5,501–10,000 | 9 |
| 10,001+ | 10 |
The general rule for planned encounters: creatures should be within two levels of the dungeon level. A 3rd-level dungeon works with creatures of levels 1–5. Creatures more than two levels above the party's expected level represent potential TPK territory; below that range, they're minor resource drains at best.
The word "expected" does some work in that sentence. A party with clever tactics, strong positioning, and lucky dice regularly defeats creatures significantly above their expected level. A party that has burned most of its resources, is split up, or has stumbled into a bad tactical situation, loses to creatures well below it. Creature level is calibration, not a guarantee. The dungeon should contain a few encounters that could kill the party under bad conditions — that's what makes the dungeon feel dangerous rather than pre-cleared.
For wilderness encounters, the calibration is looser. The wilderness is not leveled. A 1st-level party roaming the Emerald Plains can stumble across a troll den. What changes is the context: the wilderness gives the party escape routes, terrain advantages, and time to recognise a threat and disengage, that a dungeon corridor does not. If a low-level party would normally encounter something that would annihilate them, give them a reason to see it from a safe distance and choose. Let them choose. The option to flee is not a design failure — it's player skill.
BUILDING DUNGEONS
The Known World has more dungeons than it acknowledges publicly.
Second Era ruins are the most common: structures built during and before the rule of the Darkness, abandoned or destroyed when the Second Era ended. They range from collapsed towers in the Emerald Plains to intact complexes in the Dead Lands, sealed when their builders vanished. Some were fortresses. Some were temples. Some were repositories for things that should have been destroyed. Most have been partially stripped already — the Known World is 180 years old, and people have been recovering things from old ruins for most of that time. What remains tends to be things that survived the stripping because they were hidden, trapped, dangerous, or all three.
Cadman installations — military outposts, supply caches, inquisition holds — function as dungeons when the party is on the wrong side of them. These aren't ruined. They're staffed, maintained, and actively hostile. Infiltrating one is a different kind of challenge than exploring an ancient ruin: the obstacles are people with routines, not traps set once and left.
Criminal and mercenary holds occupy a middle category. A guild's underground vault, a Crimson Company safehouse, the private cellar of a merchant who imports things no one is supposed to know about — these are dungeons with living, motivated populations who know the layout far better than the party does.
Stocking a Dungeon
The traditional ratio holds: roughly one-third of rooms contain monsters or creatures, one-sixth have traps or tricks, one-sixth have unguarded treasure, and one-third are empty. That last category is the one GMs most often collapse in the name of efficiency, and collapsing it is a mistake.
Empty rooms serve multiple functions. They give the party time to breathe between meaningful encounters. They establish that the dungeon is a real place, not a sequence of fights. They contain evidence of what was here before: drag marks, scorch patterns on the stone, equipment that was dropped in a hurry, and a campfire that was burning within the last few days. A well-designed empty room isn't absent of content — it's absent of immediate threat, while containing information that makes the rest of the dungeon legible.
Keying Rooms
A room key needs four things: a number (for navigation), a brief description of what the party perceives when they first enter, the room's contents (creatures, objects, traps, exits), and any notes about how the room interacts with other rooms. The description should be written for what a party without any special knowledge or abilities would perceive — sight, sound, smell. What they perceive, not what the GM knows, is there.
For creatures, the key should note the creature's current state: sleeping, eating, on guard, or in the middle of a dispute with its companions. Creatures the party encounters while they're doing something are more interesting and more memorable than creatures standing in a room waiting to fight. An orc patrol that the party catches mid-argument has a reaction roll that reflects the interruption, not just the ambient hostility of orcs.
Connections between rooms matter. A dungeon where nothing that happens in one room affects any other is a sequence of isolated encounters, not a system. Design it as a system. When the party kills the sentries at the entrance, the creatures deeper in should have some way to find out, some chance of being warned, some mechanism for the alarm to spread. The party's actions should propagate through the dungeon the same way sound does.
Dungeon Layout
A dungeon where every room can only be reached in sequence is a corridor with furniture. The party has no navigation choices — only the order in which they face what the GM prepared. That produces attrition, not exploration.
Loops are the basic unit of interesting dungeon architecture. A loop is any arrangement of rooms and passages that gives the party more than one route to the same destination. The simplest loop is three rooms connected by two corridors — entering through one and exiting through another. Players can circle, flank, retreat, and regroup, choose which threats to confront and which to avoid entirely.
Designing with loops means asking, for any significant area: can the party approach this from more than one direction? If the answer is no, add a second path. It can be difficult to find — a collapsed passage that requires clearing, a secret door, a route through a lower level that comes up behind the main defense. Difficulty of access is different from absence. A guarded room that can be bypassed entirely through a known secondary route forces genuine tactical decisions. A guarded room that must be entered from one direction and cannot be circumvented forces only one.
Multiple entrances to the dungeon itself matter for the same reason. A single entrance is a bottleneck: the party enters once, the whole dungeon knows approximately where they are, and every return trip from town re-alerts the same positions. A second entrance — a collapsed section the party can excavate, a river access point, a sally port in a Cadman installation that opens from outside — gives the party strategic options and makes the dungeon feel like a structure rather than a funnel.
The failure mode to avoid is a linear sequence of chokepoints, where the party cannot advance until they clear each room in order, and clearing each room in order is the only way forward. This structure makes the dungeon feel scripted. The party isn't making decisions — they're working through a checklist. Design the dungeon so the party can get lost in it, go around something they're not ready for, and find something they weren't looking for. Those are the experiences the dungeon is supposed to produce.
A Note on the Dead Lands
Dungeons here are not abandoned in the conventional sense. The Second Era's end did not simply stop the inhabitants — in many cases, it preserved them in states that are neither life nor quite death. A Dead Lands dungeon should feel like a place where time stopped mid-sentence. The stocking logic is the same, but the atmosphere is different: what the party finds are not the remnants of people who left, but the remnants of people who didn't. The Darkness's influence still saturates these places. Unprotected fire behaves strangely. Mirrors show rooms as they were during the Second Era. Creatures encountered here are wrong in ways that go beyond their statistics.
RANDOM ENCOUNTER TABLES
Random encounters serve two purposes. The first is genuine procedural danger: the party lingering in a dungeon or moving through wilderness is not doing so in a stopped world, and the encounter roll makes that real. The second is generating the unexpected — the creature that wasn't in the key, the complication the GM didn't plan, the information source the party doesn't anticipate.
Build encounter tables that produce interesting combinations of creature type, motivation, and state. A table that rolls sixteen different types of monsters and nothing else produces resource drain without texture. A table that includes a group of exhausted miners who are clearly lost, a creature that appears to be following the party from two rooms behind, and a fire that is burning in a room that has no fuel produces situations.
The timing of encounter checks follows terrain. In dungeon environments, check every other turn (20 minutes) of exploration and every round of combat or loud activity. Wilderness timing varies:
| Terrain | Checks per Day | Active Hours |
|---|---|---|
| Plains | 2 | Dawn, midday |
| Scrub and brush | 3 | Dawn, midday, dusk |
| Forest | 6 | All six 4-hour watches |
| Hills | 3 | Midmorning, midday, dusk |
| Mountains | 4 | Dawn, midmorning, midday, dusk |
| Swamp | 6 | All six watches |
| Coastline | 2 | Midday, midnight |
A roll of 1 on a d6 typically indicates an encounter; modify the die type and number of rolls based on how densely populated a region is. The Cadman-occupied roads around Norderenn check more often and produce more patrols. The edges of the Dead Lands check more often and produce worse things.
Encounter size is determined after the encounter type, using a roll specific to the creature. Two orcs travelling together is a scouting pair. Twelve orcs travelling together is a warband. The first encounter offers negotiation; the second probably doesn't. Roll the number, then consider what that number implies about the creatures' purpose.
WILDERNESS DESIGN
The wilderness between locations has its own geography, ecology, and political history. The party moves through it accordingly — as travellers in a place with texture, not as combatants advancing through a procedurally generated space.
The Known World's regions each have distinct characters that should be felt during travel, not just described in paragraph-long scene-setting.
The Emerald Plains are agricultural and heavily used, which means roads, farmsteads, and the particular kind of danger that comes from places where people live close together: bandits, political entanglements, refugees from Norderenn moving south, and Cadman checkpoint infrastructure. The wilderness here is farmland between towns. Encounters are as likely to involve people as creatures.
The forests around Norderenn are occupied in the literal sense. The Cadman administration maintains garrisons and checkpoints on the main roads and treats forest travel as either smuggling or resistance activity. Moving through these forests means navigating both the natural environment and a military presence that has been in place for fifteen years. It knows the ground. The party does not.
The mountains that rim the Inner Sea's northern edge — where the Vithurian Mines run, and Korthak Citadel overlooks the passes — are older territory. The Stone Raiders have worked these approaches for generations. Whatever lives in the high passes has been doing so longer than that.
The Dead Lands require different preparation than any other wilderness region. The baseline encounter table is weighted toward remnants of the Second Era: creatures that are not alive in any normal sense, artifacts that are unsafe to touch without understanding what they are, and environmental effects that occur nowhere else. The Chroniclers maintain controlled access to the Dead Lands and are not gentle about unauthorized expeditions — an encounter with a Chronicler patrol in the Dead Lands is a serious political complication, not just a combat.
Point-of-interest design. Hex-by-hex mapping works for some tables and some players. For groups that find dense spatial accounting a barrier, the point-of-interest approach is more accessible without sacrificing the important property of wilderness — that it is traversable by decision. The GM establishes known locations, the distances and difficulties between them, what the party knows in advance about the route, and what they might encounter. The party navigates between points, and encounters happen on the route. The wilderness still has content and danger; it doesn't require centimetre-precise mapping to deliver that.
Whichever approach the table prefers, the wilderness should have landmarks. Travellers navigate by what they can see, and the party should be doing the same: the distinctive ridge that means they're two days from the waystation, the dried riverbed that runs northwest toward the Cadman checkpoint they're trying to avoid, the ruins on the hilltop that every local knows about but no one has entered. These give the party information to work with and decisions to make about their route.
TRAPS AND TRICKS
A trap is only interesting if the party could have found it.
That is the central principle. An instant-death trap with no warning is an arbitrary penalty. The party died because they stepped somewhere, not because of anything they chose. Frustration and tension are different experiences — one ends engagement, the other intensifies it.
A fair trap has detectable signs. It might be the slight difference in the stonework, the absence of dust in a straight line across the floor, the sound that stops when the party gets close. Players with no relevant skills should have a chance of spotting something wrong with careful attention. Players with applicable skills — rogues, rangers, characters with specific non-weapon proficiencies — should have a meaningful advantage. The trap is hidden in the room; the party is capable of finding it if they look.
The signs don't need to be obvious. A tripwire at ankle height in a dark corridor is not obvious, but discoverable — the party member who says "I move slowly and check the floor before each step" finds it. The party member who charges ahead does not. Player choices should determine who gets hurt, not random chance disconnected from anything anyone did.
The exception is the trap that announces itself before it triggers. A pressure plate that clicks when someone steps on it, giving a half-second before the mechanism fires, is fair even if the party didn't detect it in advance — because finding it in the moment, understanding what just happened, and deciding what to do in the next second is itself a meaningful experience. Many classic traps work this way: they're not really about preventing the trigger, they're about what happens after.
Tricks differ from traps in that they are not hostile. A room that fills with water when the wrong door is opened is a trick — the party gets wet, potentially loses equipment, and has to solve the drainage problem, but the room is an obstacle rather than a weapon. Sliding staircases, one-way passages, illusory floors that are real but appear absent, statues that point in false directions — these challenge navigation and problem-solving. They work best in variety with traps: if everything dangerous looks exactly like everything else, the party eventually treats all unusual features as threats and loses the exploratory relationship with the dungeon.
Award XP for overcoming traps — whether by disabling them, bypassing them cleverly, or triggering them and surviving. The party took a risk by being there and dealt with the result; that deserves reward. Don't award XP for traps the party avoided without knowing they existed.
Categories of traps the Known World produces. Second Era ruins contain magical traps as often as mechanical ones: wards that respond to alignment, triggered illusions designed to mislead rather than harm, extraplanar entities bound as guardians by 180-year-old agreements that are still honoring their contracts. Mechanical disabling does not work on a bound elemental; finding the binding's weakness, or offering the entity something it wants, might.
Cadman installations use professional military trap design: crossbow mechanisms on delayed triggers, pressure-activated alarm systems tied to bells two rooms away, the kind of practical engineering that a professional army has refined over generations. The Church of Tar'amaan's holdings favour detection and interrogation over elimination: a room that tells the guards who has entered is more useful to an inquisitorial operation than one that simply kills intruders. The trap categories are evidence of who built the place and why — and can function as information encounters before the party triggers anything.
REACTION ROLLS AS NARRATIVE TOOLS
The reaction roll is the mechanical acknowledgment that the world is not pre-sorted.
Most creatures in the Known World are not waiting to fight the party. They have their own purposes, and encountering a group of armed strangers is a situation they assess on its merits — who these strangers are, what they want, what the risks of confrontation look like, whether there's a better outcome available. The reaction roll captures that assessment before the party has a chance to act.
Roll 2d10 and apply the Charisma reaction modifier of whichever character speaks first or is most visible. The result:
| 2d10 | Initial Reaction |
|---|---|
| 2–3 | Hostile — attack is likely and imminent |
| 4–6 | Unfriendly — suspicious and guarded, looking for reasons |
| 7–9 | Neutral — observant but uncommitted |
| 10–12 | Indifferent — not particularly interested in the party |
| 13–15 | Friendly — open to engagement and communication |
| 16–20 | Very Friendly — actively well-disposed, willing to assist |
Hostile means the creature has decided the party is a threat or an obstacle, is ready to act on that decision, and needs only a trigger or the right moment. A hostile patrol that spots the party at a checkpoint is going to demand identification and escalate if the answer is wrong. A hostile predator that has cornered the party is assessing the risk of the attack, not waiting for permission. Violence is imminent and likely — it is not guaranteed to start immediately.
Friendly means the creature is genuinely open, willing to engage, and well-disposed to the party. A friendly guard is glad to talk, might share local knowledge, might warn the party about something ahead, and is still doing their job. A friendly bandit hasn't abandoned banditry; they're simply in a mood where robbing this particular group feels less appealing than conversation. The disposition is real. The conditions under which it holds are not unconditional.
The reaction roll is a starting point. The GM runs the NPC from that starting point based on what the party does next. A neutral reaction shifted friendly by a generous gift becomes something the NPC remembers. A friendly reaction shifted hostile by an insult or a perceived threat is information about both the NPC and the party's approach. These shifts are fiction, not mechanics — the GM reads the situation and adjusts.
Using reaction rolls in the Known World, specifically. The Known World's political texture means that initial reactions often carry faction context. A patrol soldier's reaction to the party is partly personal and partly institutional — a Cadman soldier who rolls friendly is still going to ask for the party's papers, because friendly Cadman soldiers do that. A Church of Tar'amaan inquisitor who rolls neutral is still evaluating whether the party contains magic-users, because that is their function. The reaction roll establishes emotional disposition; the creature's role establishes what they do with that disposition.
The Chroniclers are a special case. Their reaction to unauthorized Dead Lands expeditions is consistently hostile, regardless of roll — that is policy, not mood. The reaction roll, for a Chronicler patrol, determines how much latitude they offer before enforcement: a hostile Chronicler kills first; an unfriendly one demands compliance and arrests; a neutral one is watching and reporting; a friendly one might look the other way for someone they know, under specific conditions. This is the range of outcomes inside an institutional framework, and it's a better model for how factions operate than treating all members of an organisation as identical NPCs.
Random encounters generated during wilderness travel should always have a reaction roll, even when the creature type is inherently aggressive. An ogre that rolls neutral is an ogre who spotted the party but is currently more interested in the deer carcass it's been working on. This creates texture that a standing rule of "all ogres attack on sight" cannot. The party might get past without incident. They might not. The dice and the fiction together determine which.
Reaction rolls in social encounters. Chapter 12 covers the mechanics of reaction rolls in the context of planned encounters and first meetings. The principle extends to negotiation: when the party attempts to shift an NPC's reaction through conversation, gifts, or appeals to interest, the GM adjudicates whether the attempt is credible — and if it is, decides whether to shift the reaction outright or call for a Charisma check to determine the degree of shift. The reaction roll is not a locked number. It's where the encounter starts.