GM Techniques · Core Rules

Philosophy

Cerberus hides no rules from the GM: the entire system rests on three rolls and one judgment. That judgment must be reproducible—fast and consistent from one session to the next, without inventing a new rule for every question.

“Does It Exist?”

Players ask this every session: is there a stable behind this inn, a rope in this warehouse, a healer in this village? There is no single right answer.

A GM who always agrees lets players write the world in their place. One who always refuses kills their inventiveness and teaches them to stop proposing anything. The right balance is neither extreme: it means calibrating player imagination.

The tool already exists: the prescribed chance roll in 1. The GM does not answer the question directly, but converts uncertainty into a probability, announces it, and lets the die decide the world.

Player: Does this inn have a stable with horses? GM: Probably. One chance in two: roll 1 on a d2 and the stable is there.

Three safeguards apply, identical to those in 1:

Do not roll when the answer is obvious. An inn on a trade road has a stable; it is there without a die. A mountain hermitage does not. Roll only for genuine uncertainty, or the GM hides behind chance instead of describing the world.
A good argument improves the odds. “Horses were tied outside, so there is probably a stable” deserves better than one chance in two. A player's reasoning about the world matters as much as their plan for an action.
Announce the odds before rolling. After the roll, the answer becomes canon. Record it; it remains true for everyone throughout the campaign. A world decided by dice remains coherent as long as nothing is undone.

Failure does not punish. “There is no stable” is the complete answer. It does not mean “there is no stable and the innkeeper distrusts you.” The player asked a question; they did not attempt an action.

The GM remains neutral without becoming a wall, and players no longer feel that the setting is dictated to them. They discover it, which is the same process viewed from the other side of the screen.

Social Interactions

Cerberus has no social skills. No Diplomacy, Bluff, or Disguise: the prescribed chance roll replaces them all (see 1). The reason is direct—a social skill invites the player to portray a character minimally, then roll for the rest. Cerberus wants the opposite.

Yet characters are not socially equal. What distinguishes them is not a roll modifier, but how the GM makes NPCs react. A barbarian with +8 Strength does not need an Intimidation roll. They need the GM to portray NPCs who see a colossus enter the room.

Adjust for attributes. Read the sheet and adapt NPC behavior:

Attribute	High	Low
Strength	difficult to impress, respectful tone, asked for demonstrations	patronized, capabilities underestimated
Dexterity	grace admired, entrusted with delicate tasks	concern whenever precision is required
Constitution	endurance respected, physically challenged	health concerns, offers of rest
Intelligence	complex vocabulary, opinions sought on difficult subjects	simplified explanations, extra patience, underestimation
Wisdom	advice sought, moral judgments entrusted	distrust on important decisions, everything explained
Charisma	immediate attention, open tone, rapid trust	greater effort needed to persuade, neutral or wary reception

Adjust for background. A character's origin opens and closes doors:

Origin	Effect
Noble	titles respected, conduct expected, easier access to circles of power
Humble	kinship with commoners, distrust from elites
Military	respect from other warriors, direct language, strategies valued
Religious	reverence from believers, skepticism from others, wisdom expected
Criminal	underworld contacts, distrust from authorities, street codes recognized
Foreign	curiosity or suspicion depending on context, cultural misunderstandings
Local	customs understood, easy connections, loyalty expected

Five concrete levers. To apply these adjustments quickly, change an NPC's tone, vocabulary, body language, level of trust—how much information they share—and what they offer or ask of the character.

Three practices make this sustainable throughout a campaign: prepare flexible NPCs capable of several reactions; remain consistent from one session to the next; and tell players that their attributes work this way, so they can lean into them rather than search for a roll. Finally, give everyone their scenes. A low attribute is a color to play, not a permanent punishment.

Building Martial NPCs

Player characters earn most of their power organically. An ingredient changes their Strength, two months with a teacher grant a feat, and a weapon carries the history of how they found it. A newly written NPC has lived through comparable years, but the GM does not need to reconstruct every month before using them.

Build a martial NPC from five visible decisions:

Choose class and level. These set HP progression, Saves, Speed, and the Action Board. They describe professional experience, not the whole person.
Assign attributes from the NPC's life. A Level 7 soldier need not possess enormous Strength. A giant, altered barbarian, veteran archer, and aging swordmaster can reach the same level through very different bodies.
Give them believable equipment. Rank, wealth, culture, and employer determine the weapon and armor. Do not grant an enchanted weapon merely because the NPC needs a number; if they own one, know where it came from.
Choose two or three techniques that will matter in the encounter. A shield guard, mounted hunter, grappler, and assassin should play differently even when their totals are close.
Add a damage progression when training must carry the profile. Basic Fighting Style gives a professional NPC reliable weapon damage according to class and level. Use it when the NPC lacks a stronger named style. It is especially useful for a veteran whose threat comes from experience rather than extraordinary Strength, transformation, or a famous weapon.

Basic Fighting Style is a feat on the NPC's sheet, not a free class feature. A player can earn it through training under the normal organic progression rules. An NPC may instead possess a personal style when their history supports one; the personal style replaces the basic bonus rather than stacking with it. Spellcasting classes have lower grades, but can still receive the feat when the NPC's history includes serious weapon training.

Before play, compare one ordinary attack to the party's DR. If every blow disappears, decide whether that is the intended story. A poorly equipped soldier may truly be unable to harm an armored hero. A martial boss intended to remain dangerous needs a plausible answer: greater Strength, a better weapon, a fighting style, grappling, or control of the terrain. These answers are not interchangeable, because each creates a different fight.

Building an Encounter

A good encounter need not push players until all their ingenuity, strategy, and luck become essential. Easy fights belong in a campaign. They confirm that characters have grown: bandits who once posed a threat now fall to an area spell or a few well-placed attacks. If every opponent grows with the heroes, advancement becomes invisible.

Difficulty is not the starting point. Begin by imagining what is happening in the world: who is present, what they want, how they are armed, where they chose to fight, and what they will do if events turn against them. Only once the scene is visible should the GM estimate the numbers needed to make it work.

Imagine Before Balancing

Two bands of thieves have no reason to look alike. One carries bows and knives, another crossbows and swords. Highwaymen may have two riders. Street thieves might wield poisoned daggers while three accomplices fire hand crossbows from windows.

A few details are enough:

their objective—to rob, kill, capture, delay, or frighten;
their equipment and fighting style;
two or three interesting positions on the battlefield;
their discipline and the point at which they flee;
a memorable face, voice, or behavior.

These choices make the encounter memorable before the first number appears.

Estimating Power

No formula perfectly predicts a Cerberus group's actual strength. Characters grow organically, know different spells, and approach each session with different levels of caution. Make a quick estimate, then learn from what actually happens at the table.

Six Level 1 bandits, for example, may challenge three Level 2 or 3 characters. They become more dangerous if the players are rusty, spread out poorly, or swagger into an obviously prepared position. Conversely, one well-placed area spell may end the encounter.

Judgment is a muscle. After every fight, remember what truly mattered: DR, action count, Reflex, an area ability, dominant terrain, or a weakness the players exploited at once. The next estimate will be better.

Counting Damage Upward

When an opponent's proper HP total is difficult to predict, the GM need not set it immediately. Simply add damage taken—38, then 71, then 119—instead of starting from an uncertain maximum and subtracting. After seeing a few representative player attacks, choose an appropriate threshold and announce death when an opponent reaches it.

This technique is especially useful for a group of comparable creatures. Once the first bandit dies, that total normally becomes the benchmark for other bandits of the same constitution. Do not then move the threshold merely to negate a spectacular critical or arbitrarily save an opponent. Adjustment exists to understand power that preparation could not predict, not to take away the consequences of player actions.

For a boss, the GM can wait a few exchanges before locking the total. Once the boss's condition has been clearly described—gravely wounded, nearly fallen—the numbers and the narrative must continue to answer each other honestly.

Model 1 — Bandits

Fighting a numerous but weaker group is one of the game's most useful models. It lets the GM play tactically and gives players room to use multiple attacks, area spells, and powers that would overwhelm a single target.

The encounter usually follows four stages:

The GM prescribes the chance that someone in the group notices the threat.
A short roleplaying scene occurs if the bandits speak, threaten, or negotiate.
Positions are established and combat proceeds normally.
A new scene begins if bandits flee, surrender, or are captured.

Ordinary bandits usually have no magic. They may still have experience fighting casters: spreading out against area spells, shooting from cover, interrupting a casting, using smoke, or sending a fighter into melee. Experience makes them interesting without artificially granting supernatural powers.

Numbers are their main strength. The coordinated groups from 8 divide naturally by position: archers, melee fighters, and those circling the players need not share the same Reflex roll.

Model 2 — The Boss

A boss confronts the group with a character more powerful than any individual member. The players retain one fundamental advantage: together, they possess more actions, reactions, and perspectives. They can surround the boss, exhaust defenses, and create openings for one another.

A martial boss is often best for the first encounter of this kind. Their intentions remain readable: charge, strike, parry, seize, or knock down. The GM learns to manage a superior creature without simultaneously consulting a long spell list. More complex bosses can follow once the table understands this rhythm.

The performance does not end when weapons come out. A boss may speak before the fight, react to player techniques during it, and reveal something in defeat. Even an opponent destined to die benefits from remaining a character until the final moment.

Model 3 — Counterparts

Counterparts are a group of opponents built like true characters, with distinct classes, feats, spells, and roles. A party of barbarian, rogue, and wizard might face a Fighter, Cleric, and Inquisitor. They need not be exact copies; the interest lies in a different combination of strengths.

This model requires more preparation and remains the hardest to estimate, especially at high level. The GM has two reliable levers:

create deliberately weaker opponents, then use their abilities fully;
give them power close to the heroes', but play them with mistakes, hesitation, bad information, or objectives that prevent them from optimizing every action.

The second method is not cheating: opponents are characters. They may panic, protect an ally, preserve a spell for later, or misunderstand the players' strategy. In complex high-level combat, nobody plays perfectly—not the players, not the GM—and the system does not require it.

Model 4 — The Ambush

An ambush uses the ordinary rules, but begins with a hidden threat and a violent advantage for the attacker.

First prescribe the chance that a group member notices the danger. On success, the scene may become a contest of position or a normal fight. On failure, attackers strike from hiding under the Perception rules (2). If several characters are equally exposed, the GM may choose the first target at random.

If the failed detection leaves no plausible time to react, the hidden attackers act first and their opening attack actions are automatic critical hits. Otherwise, use the ordinary Reflex duel for attacking from hiding; a quick defender may remove the critical and act before the ambushers.

An ambush need not be balanced like a frontal battle. It asks a different question: can the characters absorb the first shock, find cover, and regain control before the attackers exploit their advantage?

Controlling Power Growth

This is an eternal GM problem. Casters are the most powerful characters in the game, and that is not a flaw: Cerberus does not balance every race and profile against the others (see 12). Yet a caster's rise to power should take time and be earned.

The recommended approach:

Advance levels slowly. Level governs general capabilities and need not increase often.
Grant no spell automatically on gaining a level. An important new spell demands study time within the game—this binds roleplay to gameplay instead of advancing them on parallel tracks.
Separate caster level from character level. The GM grants mana power without granting a level.

CL governs the magnitude of spells already known, while level opens complex spells and the action economy (see 5). Raising CL alone lets the player feel progress—spells hit harder, last longer, and reach farther—without adding any new spell or invalidating written scenarios. The sensation without the leap.

Where Are the Other GM Notes?

The remaining techniques live in the chapters they support:

Technique	Chapter
The three golden rules of prescribed rolls	Base Resolution
Converting a hazard into odds and damage	Environment and Physical Hazards
Access to rest sets difficulty	Rest and Recovery
Distributing magic items	Enchanting Magic Items
Surprise and attacking from hiding	Perception and Stealth

Chapter Summary

Resolve a question about the world like an action: announce the odds, then roll.
Do not roll when the answer is obvious; a good argument improves the odds.
What the die decides becomes canon and is not undone.
Failure answers “no”; it adds no consequence.
There are no social skills: NPC behavior changes with a character's attributes and background, rather than through a roll modifier.
An encounter need not be difficult. It must first belong in the fiction and reveal what the characters have become.
Bandits, bosses, counterparts, and ambushes are four reliable structures the GM can reskin and reuse.
To advance a caster without unbalancing everything, raise caster level without raising character level.