Weighted Random Selection Without Bias

Description: How to implement rarity systems, loot tables, and weighted outcomes without introducing hidden probability errors.

The Problem

Many real-world systems require outcomes that are not equally likely.

Example:

• Common item: 70%
• Rare item: 25%
• Legendary item: 5%

Developers often implement this using quick heuristics that silently distort probabilities. This leads to:

• Players distrusting loot systems
• Regulatory issues in gambling environments
• Long-term economic imbalance in games

Weighted randomness must be implemented precisely, or the distribution drifts over time.

What Is Weighted Randomness?

Instead of selecting uniformly from N outcomes, each outcome has a weight.

Example:

• Common = 70
• Rare = 25
• Legendary = 5

Total weight: 70 + 25 + 5 = 100

Probability of each:

• Common = 70 / 100
• Rare = 25 / 100
• Legendary = 5 / 100

The goal is to select an outcome proportional to its weight.

The Most Common Incorrect Implementation

Typical mistake:

1. Generate random integer 0–100
2. Map ranges manually

Problems:

• Off-by-one errors
• Gaps or overlaps
• Future edits break distribution
• Hard to audit

Even small mistakes produce measurable bias over thousands of draws.

Correct Method: Cumulative Distribution

Step 1 — Define weights

Common = 70 Rare = 25 Legendary = 5

This guarantees perfect proportionality.

Why This Matters in Production Systems

Loot Economies

A 1–2% drift:

• Inflates rare item supply
• Breaks marketplace pricing
• Causes long-term economy collapse

Gambling & Compliance

Weighted outcomes must be:

• Reproducible
• Auditable
• Verifiable

Incorrect implementations can:

• Fail certification
• Trigger legal issues

Live Service Games

Small bias × millions of events =

• Player suspicion
• Reddit investigations
• Reputation damage

Most “rigged RNG” accusations come from bad weighted selection.

Common Engineering Mistakes

Using Floating Point

Float rounding causes:

• Probability drift
• Platform-dependent results

Use integers only.

Renormalizing Every Draw

Changing total weight dynamically:

• Breaks expected distribution
• Makes audits impossible

Editing Tables Without Versioning

Changing weights without:

• Version IDs
• Seed domain separation Makes past results unverifiable.

Deterministic Auditability Pattern

Inputs:

• Player seed
• Server seed
• Public entropy
• Counter

Process:

1. Hash inputs
2. Derive 64-bit integer
3. Map via rejection sampling
4. Select from cumulative weights

Result: Anyone can recompute the exact selection.

Key Takeaway

Weighted randomness is not just picking from ranges. Correct implementation requires:

• Uniform source entropy
• Bias-free range mapping
• Integer arithmetic
• Deterministic replay

If any step is wrong, the distribution becomes statistically provable as unfair.

Where This Is Used

• Loot boxes
• Card rarity
• Prize wheels
• Ad allocation
• Match rewards
• Raffles with tiered prizes
• On-chain games
• Provably fair casinos

Weighted selection is one of the most economically sensitive parts of any randomness system.