123 Million CS2 Simulations: Engineering Reliable Weighted RNG

What 123 million simulated CS2 case openings taught me about modeling RNG

Graysonwerner analyzed 123 million simulated Counter-Strike 2 case openings through case-sim.com to identify subtle RNG bugs in probability code. The study highlights how Valve’s disclosed 2017 drop rates reveal edge cases in standard loot table implementations.

Why This Matters

Idealized probability models often fail at scale due to IEEE 754 floating-point precision errors, where sums like 0.7992 + 0.1598 + 0.0320 + 0.0064 + 0.0026 result in 0.9999999999999999 instead of 1.0. This technical discrepancy can cause fall-off-the-end bugs in high-traffic systems like loot boxes or A/B testing frameworks, leading to skewed distributions or system crashes when Math.random() returns a value exceeding the cumulative float total.

Key Insights

• Integer-based weights (e.g., parts-per-10,000) are superior to float weights for ensuring probability invariants sum to exactly 1.0 in JavaScript, as seen in the case-sim.com refactor.
• Valve’s 2017 disclosed tiers (Mil-Spec to Knife/Glove) require strict enforcement of a 100% sum invariant in unit tests to prevent logic drift.
• The Spin Animation Bias Trap occurs when UI position determines results; robust systems must roll the outcome first and then map the animation to the result.
• Skin-specific float ranges, such as the AWP Asiimov’s 0.18 minimum, create artificial scarcity that uniform 0.0 to 1.0 distributions fail to model accurately.
• Pattern-based variants like Dopplers (Ruby/Sapphire) must be modeled as distinct items with specific weights to avoid over-representing rare phases.

Working Examples

Integer-based weighted RNG to avoid floating-point precision errors.

const TIER_WEIGHTS_INT = { milSpec: 7992, restricted: 1598, classified: 320, covert: 64, rare: 26 }; function rollTier() { const r = Math.floor(Math.random() * 10000); let cumulative = 0; for (const [tier, weight] of Object.entries(TIER_WEIGHTS_INT)) { cumulative += weight; if (r < cumulative) return tier; } }

Correctly calculating wear values using skin-specific workshop float constraints.

function rollFloat(skin) { const raw = Math.random(); return raw * (skin.maxFloat - skin.minFloat) + skin.minFloat; }

Practical Applications

• Loot reveal UI in Gacha systems: Roll the result first rather than deriving the item from the visual stop position to avoid skewing actual drop rates.
• High-frequency event tracking: Design schema around aggregate counters per item and day rather than writing individual rows to prevent database write exhaustion.
• Probability validation in CI: Implement unit tests that assert tier weights sum to an exact integer (e.g., 10000) to catch copy-paste errors in case definitions.

References:

• https://dev.to/graysonwerner100commits/what-123-million-simulated-cs2-case-openings-taught-me-about-modeling-rng-7dh