Randomness & Procedural Generation

Randomness keeps games fresh: loot drops, enemy spawns, critical hits, and entire levels in roguelikes and Minecraft. But “random” in games is a design tool, and raw Math.random() is rarely what you want unshaped.

Seeded randomness

A pseudo-random number generator (PRNG) produces a sequence that looks random but is fully determined by its seed. Same seed → same sequence → same level. That gives you shareable level codes (“seed 1337”), daily challenges where everyone plays the same map, reproducible bug reports, and replays.

const rng = seeded(1337)     // a tiny PRNG such as mulberry32
rng() // 0.2147…  always the same first number for seed 1337
rng() // 0.8836…

Procedural generation

Procedural content generation (PCG) builds content with algorithms. The classic recipe: start with noise, then apply rules to shape it into something meaningful.

for (let step = 0; step < STEPS; step++)
  grid = grid.map((row, y) => row.map((cell, x) => {
    const walls = countWallNeighbours(grid, x, y)    // 0..8
    return walls >= 5 ? WALL : walls <= 3 ? FLOOR : cell
  }))

Shaping randomness for fun

• Weighted tables: common 70%, rare 25%, legendary 5%.
• Shuffle bags: put items in a bag, shuffle, draw until empty, refill. Tetris does this so you never wait 30 pieces for a line piece.
• Pity timers: increase the chance after each failure, so bad luck streaks have a limit.
• Random ranges, not raw random: speed = 100 + rng() * 20 adds variety without chaos.

function weightedPick(table, rng) {     // [{ item: 'sword', weight: 5 }, …]
  let r = rng() * table.reduce((sum, e) => sum + e.weight, 0)
  for (const e of table) if ((r -= e.weight) < 0) return e.item
}