Simulation Config

Simulation Configuration

Configure TraitorSim game parameters to test different regional rule variants, recruitment mechanics, end-game scenarios, and player counts. This skill helps set up simulations that match specific franchise rules or test custom configurations.

Quick Start

from src.traitorsim.core.config import SimulationConfig

# UK rules (standard)
uk_config = SimulationConfig(
    rule_set="UK",
    recruitment_type="standard",
    end_game_mode="vote_to_end",
    tie_break_method="revote",
    num_players=22,
    num_traitors=4,
    starting_pot=0,
    enable_dramatic_entry=True,
    shield_visibility="secret"
)

# US rules
us_config = SimulationConfig(
    rule_set="US",
    recruitment_type="standard",
    end_game_mode="vote_to_end",
    tie_break_method="random",
    num_players=20,
    num_traitors=3,
    enable_dramatic_entry=False
)

# Australia rules (Traitor's Dilemma)
aus_config = SimulationConfig(
    rule_set="Australia",
    recruitment_type="ultimatum",
    end_game_mode="traitors_dilemma",
    tie_break_method="countback",
    num_players=24,
    num_traitors=4,
    enable_dramatic_entry=True
)

Regional Rule Variants

UK Rules (Standard)

Player Count: 20-22 contestants Traitor Count: 3-4 initial Traitors

Recruitment:

• Triggered when a Traitor is banished
• Faithful can decline recruitment
• If declined, game continues with fewer Traitors

End Game:

• Vote to End at Final 4 or fewer
• All players vote: Continue or End?
• Requires unanimous consent to end
• If vote fails, game continues (murder happens)

Tie-Breaking (Round Table votes):

• Revote with tied players immune
• If still tied after revote: No banishment

Breakfast Order:

• Dramatic Entry enabled
• Last player to enter breakfast is often suspicious
• GameMaster can manipulate entry order for drama

Shield:

• Secret - No one knows who has Shield until used

Prize Pot:

• Starts at £0
• Missions add £1,000-£10,000
• Final pot typically £20,000-£120,000

Configuration:

SimulationConfig(
    rule_set="UK",
    recruitment_type="standard",
    end_game_mode="vote_to_end",
    tie_break_method="revote",
    num_players=22,
    num_traitors=4,
    starting_pot=0,
    enable_dramatic_entry=True,
    shield_visibility="secret"
)

US Rules

Player Count: 20 contestants Traitor Count: 3 initial Traitors

Recruitment:

• Same as UK (Faithful can decline)

End Game:

• Vote to End at Final 4
• Majority vote (not unanimous)
• If vote fails, game continues

Tie-Breaking:

• Random draw - Flip coin or draw lots
• No revote

Breakfast Order:

• No dramatic entry manipulation
• Players enter naturally

Shield:

• Sometimes public (announced at Round Table)

Prize Pot:

• Starts at $250,000
• Missions deduct from pot on failure
• Final pot typically $50,000-$200,000

Configuration:

SimulationConfig(
    rule_set="US",
    recruitment_type="standard",
    end_game_mode="vote_to_end",
    tie_break_method="random",
    num_players=20,
    num_traitors=3,
    starting_pot=250000,
    enable_dramatic_entry=False,
    shield_visibility="public"
)

Australia Rules

Player Count: 22-24 contestants Traitor Count: 4 initial Traitors

Recruitment:

• Ultimatum (Blackmail) mode
• When last Traitor remains, can force recruitment
• "Join me or I'll reveal you as the next murder victim"
• Faithful cannot safely decline

End Game:

• Traitor's Dilemma at Final 2 (if both Traitors)
• Prisoner's Dilemma: Share or Steal?
• Both Share: Split pot 50/50
• One Steals: Stealer gets 100%, Sharer gets £0
• Both Steal: Both get £0

Tie-Breaking:

• Countback - Cumulative votes across all previous Round Tables
• Player with most total votes is banished
• If still tied: Random

Breakfast Order:

• Dramatic entry enabled

Shield:

• Secret

Prize Pot:

• Starts at AU$250,000
• Missions add to pot
• Final pot typically AU$100,000-AU$250,000

Configuration:

SimulationConfig(
    rule_set="Australia",
    recruitment_type="ultimatum",
    end_game_mode="traitors_dilemma",
    tie_break_method="countback",
    num_players=24,
    num_traitors=4,
    starting_pot=250000,
    enable_dramatic_entry=True,
    shield_visibility="secret"
)

Configuration Parameters

rule_set: str

Options: "UK", "US", "Australia", "custom"

Default: "UK"

Impact:

• Sets default values for other parameters
• Affects GameMaster narration style
• Influences cultural references

Custom: Allows mixing rules from different variants

num_players: int

Range: 10-30

Typical values:

• Small game: 10-12
• Standard: 20-24
• Large game: 25-30

Impact on gameplay:

• Smaller games (10-15): Faster pace, fewer days, easier to track alliances
• Larger games (20+): Longer seasons, more complex social dynamics, harder to form unanimous votes

num_traitors: int

Range: 2-6

Typical ratios:

• 15% Traitors: 3 Traitors / 20 players
• 18% Traitors: 4 Traitors / 22 players
• 20% Traitors: 6 Traitors / 30 players

Impact on gameplay:

• Fewer Traitors: Harder for Traitors to win, more paranoia
• More Traitors: Voting bloc advantage, easier to recruit

Critical: If Traitors ever outnumber Faithfuls at Vote to End, Traitors auto-win

recruitment_type: str

Options: "standard", "ultimatum", "none"

Standard:

• Triggered when Traitor banished
• Faithful can decline
• If declined, game continues with reduced Traitors

Ultimatum (Blackmail):

• Last remaining Traitor forces "Join or Die"
• Faithful can technically decline, but will be murdered
• Adds tension to end game

None:

• No recruitment ever
• If Traitors eliminated, Faithfuls auto-win

Impact:

• standard: Balanced, allows Faithful heroes
• ultimatum: Favors Traitors in end game
• none: Favors Faithfuls

end_game_mode: str

Options: "vote_to_end", "traitors_dilemma", "elimination"

Vote to End:

• At Final 4 (or fewer), players vote to continue or end
• If unanimous to end: Remaining Faithfuls split pot
• If vote fails: Game continues (Traitors murder)

Traitor's Dilemma:

• At Final 2 with 2 Traitors: Share or Steal decision
• See Australia rules above

Elimination:

• Game continues until only Traitors OR only Faithfuls remain
• No vote to end

Impact:

• vote_to_end: Requires unanimous trust (hard to achieve)
• traitors_dilemma: Adds drama to Traitor vs Traitor end
• elimination: Longest games, highest body count

tie_break_method: str

Options: "revote", "countback", "random", "no_banishment"

Revote:

• Tied players are immune
• Re-vote on remaining nominees
• If still tied: No banishment

Countback:

• Sum all votes each tied player received across entire season
• Player with most cumulative votes is banished
• If still tied: Random

Random:

• Flip coin / draw lots
• Immediate resolution

No Banishment:

• Ties always result in no banishment
• Simpler, favors Traitors (no Traitor eliminated)

Impact:

• revote: Dramatic, requires social maneuvering
• countback: Rewards consistent voting, punishes polarizing players
• random: Fastest, but feels unfair
• no_banishment: Simplest, least drama

starting_pot: int

Range: 0 to 500,000

Typical values:

• UK: £0 (missions add to pot)
• US: $250,000 (missions deduct on failure)
• Australia: AU$250,000 (missions add)

Impact:

• Starting high + deductions: Missions feel high-stakes
• Starting zero + additions: More forgiving, cooperative vibe

enable_dramatic_entry: bool

Default: True

Effect:

• GameMaster can manipulate breakfast entry order
• Last player to enter is often suspicious ("Why were they late?")
• Creates "breakfast order tell" - Traitors may consistently enter last

Impact:

• Adds observable clues for Faithfuls
• Creates extra paranoia

shield_visibility: str

Options: "secret", "public"

Secret:

• No one knows who has Shield
• Shield holder can bluff ("I have Shield, so don't waste murder on me")
• Creates information asymmetry

Public:

• Announced at Round Table who has Shield
• No bluffing possible
• Shield holder is obvious target AFTER shield used

Impact:

• secret: More strategic, allows Shield bluffing
• public: Simpler, less deception

personality_generation: str

Options: "archetype", "random" (deprecated)

Archetype:

• Uses persona library with backstories
• Grounded in demographics and psychology
• Preferred for realistic gameplay

Random (deprecated):

• Random OCEAN trait assignment
• No backstory or demographic grounding
• Not recommended

Impact:

• archetype: Rich, emergent social dynamics
• random: Shallow, unpredictable agents

Instructions

When Setting Up a Simulation

1. Choose rule set:

   # Standard UK rules
   config = SimulationConfig(rule_set="UK")
   
   # Or custom mix
   config = SimulationConfig(
       rule_set="custom",
       recruitment_type="ultimatum",  # From Australia
       end_game_mode="vote_to_end",   # From UK
       tie_break_method="random"      # From US
   )
   

2. Adjust player counts:

   config.num_players = 15  # Smaller game for testing
   config.num_traitors = 3  # 20% ratio
   

3. Set persona library:

   config.persona_library_path = "data/personas/library/test_batch_001_personas.json"
   

4. Validate configuration:

   config.validate()
   # Raises error if invalid (e.g., more Traitors than players)
   

5. Run simulation:

   from src.traitorsim.core.game_engine import GameEngine
   
   engine = GameEngine(config=config)
   results = engine.run()
   

When Testing Rule Variants

Compare UK vs US end game:

configs = [
    SimulationConfig(rule_set="UK", end_game_mode="vote_to_end"),
    SimulationConfig(rule_set="US", end_game_mode="vote_to_end", tie_break_method="random")
]

results = []
for config in configs:
    engine = GameEngine(config=config)
    result = engine.run()
    results.append(result)

# Compare outcomes
for i, result in enumerate(results):
    print(f"Config {i+1}: Winner: {result['winner']}, Days: {result['days']}")

Test recruitment mechanics:

# Standard recruitment
config_standard = SimulationConfig(recruitment_type="standard")

# Ultimatum recruitment
config_ultimatum = SimulationConfig(recruitment_type="ultimatum")

# No recruitment
config_none = SimulationConfig(recruitment_type="none")

# Run multiple games with each config
for config in [config_standard, config_ultimatum, config_none]:
    for trial in range(10):
        engine = GameEngine(config=config)
        result = engine.run()
        # Analyze Traitor win rate

When Debugging Game Rules

Test edge cases:

1. All Traitors banished early:

   # Run game, manually banish all Traitors
   # Expected: Faithfuls auto-win
   

2. Final 2 with 2 Traitors:

   config = SimulationConfig(end_game_mode="traitors_dilemma")
   # Manually set state to Final 2 Traitors
   # Expected: Share/Steal decision triggers
   

3. Traitor majority at Vote to End:

   # Set state to Final 4: 3 Traitors, 1 Faithful
   # Expected: Traitors vote to end, auto-win
   

4. Countback tie-break with equal votes:

   config = SimulationConfig(tie_break_method="countback")
   # Create scenario where 2 players have identical cumulative votes
   # Expected: Falls back to random
   

Configuration Best Practices

For Testing/Development

Use small games:

config = SimulationConfig(
    num_players=10,
    num_traitors=2,
    enable_dramatic_entry=False,  # Simpler
    tie_break_method="random"     # Faster
)

Benefits:

• Faster execution (fewer agents)
• Easier to debug (fewer interactions)
• Quicker iteration

For Realistic Simulation

Use standard UK rules with full persona library:

config = SimulationConfig(
    rule_set="UK",
    num_players=22,
    num_traitors=4,
    personality_generation="archetype",
    persona_library_path="data/personas/library/production_100_personas.json",
    enable_dramatic_entry=True,
    shield_visibility="secret"
)

Benefits:

• Matches actual show rules
• Rich social dynamics
• Emergent behaviors

For Research/Analysis

Vary one parameter at a time:

baseline = SimulationConfig(rule_set="UK")

# Test impact of recruitment type
configs = [
    baseline,
    SimulationConfig(rule_set="UK", recruitment_type="ultimatum"),
    SimulationConfig(rule_set="UK", recruitment_type="none")
]

# Run 50 trials per config
for config in configs:
    for trial in range(50):
        result = run_simulation(config)
        save_result(result, config_name=config.recruitment_type)

# Analyze Traitor win rates

Common Configuration Errors

Error 1: More Traitors Than Players

config = SimulationConfig(num_players=10, num_traitors=15)
# ❌ ValueError: num_traitors (15) exceeds num_players (10)

Fix: Ensure num_traitors < num_players

Error 2: Invalid Rule Set

config = SimulationConfig(rule_set="Canada")
# ❌ ValueError: Invalid rule_set. Must be: UK, US, Australia, custom

Fix: Use valid rule set name

Error 3: Traitor Majority at Start

config = SimulationConfig(num_players=10, num_traitors=6)
# ⚠️  Warning: Traitors start with majority (6/10). Game balance broken.

Fix: Keep Traitors to 15-20% of total:

config = SimulationConfig(num_players=10, num_traitors=2)  # 20%

Error 4: Missing Persona Library

config = SimulationConfig(
    personality_generation="archetype",
    persona_library_path="data/personas/library/nonexistent.json"
)

engine = GameEngine(config=config)
# ❌ FileNotFoundError: Persona library not found

Fix: Generate library first or use correct path

Error 5: Invalid Tie-Break Method

config = SimulationConfig(tie_break_method="majority_vote")
# ❌ ValueError: Invalid tie_break_method

Fix: Use valid method: revote, countback, random, or no_banishment

Advanced Configuration

Custom Mission Sets

from src.traitorsim.missions.mission_registry import MissionRegistry

config = SimulationConfig(rule_set="UK")

# Override default missions
custom_missions = MissionRegistry()
custom_missions.register("Laser Heist", primary_stat="dexterity", difficulty=0.7)
custom_missions.register("Quiz Challenge", primary_stat="intellect", difficulty=0.6)
custom_missions.register("Trust Fall", primary_stat="social_influence", difficulty=0.5)

config.mission_registry = custom_missions

Seeded Randomness (Reproducible Games)

config = SimulationConfig(
    rule_set="UK",
    random_seed=42  # Same seed = same game every time
)

# Run game twice with same seed
result1 = GameEngine(config=config).run()
result2 = GameEngine(config=config).run()

assert result1 == result2  # Identical outcomes

Hybrid Rule Sets

Example: UK rules + US tie-breaking

config = SimulationConfig(
    rule_set="custom",
    recruitment_type="standard",      # UK
    end_game_mode="vote_to_end",      # UK
    tie_break_method="random",        # US
    enable_dramatic_entry=True,       # UK
    num_players=20,                   # US
    num_traitors=3,                   # US
    starting_pot=100000               # Hybrid
)

When to Use This Skill

Use this skill when:

• Setting up a new simulation
• Testing different regional rule variants
• Comparing recruitment mechanics
• Debugging game rule logic
• Creating custom configurations
• Running research experiments with varying parameters

When NOT to Use This Skill

Don't use this skill for:

• Analyzing game outcomes (use game-analyzer skill)
• Debugging agent memory (use memory-debugger skill)
• Generating personas (use persona-pipeline skill)
• Validating content (use world-bible-validator skill)