RNG System

Crimsonland uses a deterministic random number generator for spawns, bonus drops, perk selection, and various game events. Understanding the RNG helps predict and manipulate game outcomes.

Random Number Generator

From the source code references, Crimsonland uses a custom RNG implementation: RNG Type: CrandLike (C stdlib rand-compatible) Implementation: Located in grim.rand module Characteristics:

Deterministic: Same seed produces same sequence
Reproducible: Replays use same seed for parity
Fast: Optimized for frame-by-frame generation

Seed System

Seed Initialization: Game start, mode selection, or manual seed input Deterministic Replay: Survival mode replays use recorded seed to reproduce exact gameplay Seed Manipulation: Advanced players can use known seeds for specific outcomes (speedruns, challenges)

RNG Applications

Creature Spawns

Spawn Template Selection:

# Pseudocode from spawn logic
spawn_id = random_spawn_id_from_pool()
tint = random_tint() if template.tint is None else template.tint
size = random_size() if template.size is None else template.size
speed = random_speed() if template.move_speed is None else template.move_speed

Randomized Elements:

Which spawn template (SpawnId)
Creature tint (RGBA)
Creature size (0.8x - 1.5x typical range)
Movement speed (0.8x - 1.3x typical range)
Heading direction (0 - 2π radians)
Spawn position (screen edge location)

Fixed Elements (template-specific):

Creature type (Zombie, Lizard, etc.)
AI mode
Flags (split on death, ranged attack, etc.)

Spawn Timing: Random variance added to spawn timer (prevents predictable waves)

Bonus Drops

Drop Chance Roll:

# Simplified bonus drop logic
if random() < base_drop_rate:
    bonus_id = weighted_random_bonus()
    spawn_bonus(bonus_id)

Weighted Selection: Different bonuses have different spawn weights Approximate Weights (inferred from gameplay):

Bonus	Relative Weight	Rarity
Points	High	Common
Weapon	High (with Pistol)	Common
Fireblast	Medium	Common
MediKit	Medium	Medium
Shield	Low	Rare
Nuke	Very Low	Very Rare
Weapon Power Up	Low	Rare
Double XP	Low	Rare

Pistol Bias: While holding Pistol:

Overall drop rate × 1.5
Weapon bonus forced 75% of the time

Bonus Magnet Perk: Second roll on failed drop (effectively +30-40% drop rate)

Perk Selection

Selection Pool Generation:

# Pseudocode from perk selection
available_perks = filter_perks_by_mode_and_prerequisites()
shuffle(available_perks)  # RNG determines order
offered_perks = available_perks[:selection_count]  # First 5 (or 7)

RNG Determines:

Which perks appear in selection
Order of perks displayed

NOT Randomized:

Perk effects (fixed per perk)
Prerequisites (hard-coded)
Mode availability (flag-based)

Perk Expert/Master: Increases selection_count (5 → 7) but doesn’t change RNG

Combat RNG

Dodge Rolls (Dodger/Ninja perks):

if random() < dodge_chance:
    avoid_damage()
else:
    apply_damage()

Dodger: 20% chance (random() < 0.2) Ninja: 33% chance (random() < 0.333…) Poison Application (Poison Bullets):

if random() < 0.125:  # 1-in-8 chance
    apply_poison()

Jinxed Perk Events:

if random() < 0.1:
    take_self_damage(5)
if not freeze_active and random() < creature_death_chance:
    kill_random_creature()

Fatal Lottery:

if random() < 0.5:
    award_xp(10000)
else:
    kill_player()

Spread and Accuracy

Projectile Spread:

# Weapon spread affects heading
spread_offset = random_gaussian() * current_spread_heat
projectile_heading = aim_heading + spread_offset

Spread Heat: Increases per shot, decays over time Sharpshooter: Forces spread_heat to near-zero (RNG still called but clamped) Random Gaussian: Uses RNG to generate normal distribution for realistic spread

Visual Effects

Particle Systems:

Particle velocity variance
Particle lifetime variance
Particle rotation randomization
Color tint variation

Decal Placement:

Blood splatter positioning
Decal rotation
Decal scale variance

Audio Variation:

Attack sound selection (attack_01 vs attack_02)
Volume randomization
Pitch variation

RNG Manipulation

Seed Control

Known Seed Benefits:

Reproducible runs (speedruns, verification)
Favorable spawn patterns
Perk selection guarantees
Bonus drop optimization

Seed Input: Some game modes allow manual seed entry for challenges Seed Recording: Replays store seed for exact reproduction

Frame-Perfect Inputs

RNG Advancement: Each frame advances RNG state Input Timing: Slightly delayed input (1-2 frames) can change RNG state before perk selection Manipulation Strategy:

Know desired perk
Delay level-up (kill enemies slower)
RNG advances during delay
Different RNG state = different perk pool
Select perk when desired one appears

Difficulty: Requires precise frame counting and knowledge of RNG sequence

Bonus Manipulation

Bonus Spawn Timing:

Kill timing affects bonus rolls
Multiple simultaneous kills = multiple rolls
Spread kills over time for more attempts

Bonus Magnet: Doubles attempts per kill, increases RNG roll count

Determinism and Replay

Replay System

From src/crimson/replay/ directory: Replay Recording:

Initial RNG seed
Player inputs (frame-by-frame)
Game mode and settings

Replay Playback:

Restores exact RNG seed
Replays inputs identically
Reproduces exact game state

Parity: Replays match original frame-perfectly (when working correctly)

Desyncs

Causes:

Floating-point precision differences
Code changes between versions
Platform-specific RNG differences
Timer/frame rate variance

Prevention: Crimsonland uses fixed-point math and deterministic RNG to minimize desyncs

RNG Patterns

Spawn Waves

Pattern Observation: Early game spawns follow rough patterns: Minutes 0-2: Mostly Zombies and basic Lizards Minutes 2-5: Mix of Zombies, Lizards, Spiders Minutes 5-10: Aliens appear, tougher variants Minutes 10+: Boss variants, all types, high density RNG Influence: Exact timing and mix randomized, but general progression follows pattern

Bonus Clustering

Observed Behavior: Bonuses sometimes cluster (3-4 in quick succession) or drought (long gaps) RNG Explanation: Random distribution creates natural clusters and gaps Gambler’s Fallacy: Long drought doesn’t increase next drop chance (each roll independent) Actual Mechanics: Bonus Magnet provides second roll, but base rate unchanged

Statistical Analysis

Expected Values

Dodger Perk (20% dodge):

Expected hits to death at 100 HP, 10 damage/hit:
Without Dodger: 10 hits
With Dodger: 10 / 0.8 = 12.5 hits average

Ninja Perk (33% dodge):

Expected hits to death:
10 / 0.67 = 14.9 hits average

Poison Bullets (12.5% chance):

Expected hits to poison:
1 / 0.125 = 8 hits average

Fatal Lottery (50% death):

picks: 25% survival both times
picks: 12.5% survival all times
picks: 6.25% survival

Variance

High Variance Events:

Fatal Lottery (binary outcome)
Jinxed perk (random damage or benefit)
Bonus drops (long droughts possible)

Low Variance Events:

Dodge rolls (frequent, averages out)
Poison application (frequent, averages out)
Spawn timing (minor variance)

RNG Trivia

Why Crimsonland uses deterministic RNG

Deterministic RNG enables:

Exact replay functionality
Fair speedrun verification
Reproducible bug testing
Parity between original and rewrite

True random would make replays impossible.

RNG call frequency

At 60 FPS, Crimsonland calls RNG approximately:

10-20 times per frame (spawn checks, visual effects)
600-1200 times per second
36,000-72,000 times per minute

Heavy usage requires fast RNG implementation.

Speedrun seed manipulation

Speedrunners can use favorable seeds to guarantee:

Early Plasma Rifle weapon drop
Fastshot + Fastloader + Bloody Mess in first 3 levels
No boss spawns in first 5 minutes

But most categories ban manual seed input.

Perks - RNG-based perk selection
Bonuses - Bonus drop RNG
Creatures - Spawn template randomization
Survival Mode - Spawn patterns

Source Code References

grim.rand - RNG implementation
src/crimson/replay/ - Replay system using deterministic RNG
src/crimson/bonuses/selection.py - Bonus drop RNG
src/crimson/perks/selection.py - Perk selection RNG
src/crimson/creatures/spawn.py - Spawn template randomization

Overview

Game Modes

Combat System

Advanced

RNG System

RNG System

Random Number Generator

Seed System

RNG Applications

Creature Spawns

Bonus Drops

Perk Selection

Combat RNG

Spread and Accuracy

Visual Effects

RNG Manipulation

Seed Control

Frame-Perfect Inputs

Bonus Manipulation

Determinism and Replay

Replay System

Desyncs

RNG Patterns

Spawn Waves

Bonus Clustering

Statistical Analysis

Expected Values

Variance

RNG Trivia

Source Code References

Build docs developers (and LLMs) love

Overview

Game Modes

Combat System

Advanced

​RNG System

​Random Number Generator

​Seed System

​RNG Applications

​Creature Spawns

​Bonus Drops

​Perk Selection

​Combat RNG

​Spread and Accuracy

​Visual Effects

​RNG Manipulation

​Seed Control

​Frame-Perfect Inputs

​Bonus Manipulation

​Determinism and Replay

​Replay System

​Desyncs

​RNG Patterns

​Spawn Waves

​Bonus Clustering

​Statistical Analysis

​Expected Values

​Variance

​RNG Trivia

​Related Documentation

​Source Code References

Build docs developers (and LLMs) love

RNG System

Random Number Generator

Seed System

RNG Applications

Creature Spawns

Bonus Drops

Perk Selection

Combat RNG

Spread and Accuracy

Visual Effects

RNG Manipulation

Seed Control

Frame-Perfect Inputs

Bonus Manipulation

Determinism and Replay

Replay System

Desyncs

RNG Patterns

Spawn Waves

Bonus Clustering

Statistical Analysis

Expected Values

Variance

RNG Trivia

Related Documentation

Source Code References