CS2 Case Odds & Opening Statistics
What a standard weapon case can drop, how unlikely gold really is, and why probability is never a guarantee.
Last updated: by CS2Gambling.com
Published standard weapon-case rarity odds
Perfect World disclosed these historical weapon-case rates in 2017. They are a reference for standard cases, not sticker capsules, souvenir packages, terminals or custom third-party cases. Displayed historical percentages sum to 99.99999% because of rounding.
| Rarity | Per opening | Approximately 1 in | Evidence |
|---|---|---|---|
| Mil-Spec | 79.92327% | 1.25 | Published historical rate |
| Restricted | 15.98465% | 6.26 | Published historical rate |
| Classified | 3.19693% | 31.28 | Published historical rate |
| Covert | 0.63939% | 156.40 | Published historical rate |
| Rare special item | 0.25575% | 391.01 | Published historical rate |
Sources: Container history and Perfect World disclosure · Published rarity ratios explained. Neither source verifies present-day prices or precise wear distributions.
How many cases for a gold?
Gold is a knife or gloves depending on the case pool. At the historical 0.25575% rate, the mean waiting time is about 391 openings. That is not a deadline or pity timer: the next case keeps the same probability after every miss.
50% chance of at least one gold
271 cases
50% still get no gold
90% chance of at least one gold
900 cases
10% still get no gold
99% chance of at least one gold
1,799 cases
1% still get no gold
At least one gold
22.59%
No gold: 77.41%
Formula: P(at least one) = 1 − (1 − p)n, with p = 0.0025575. Assumes independent openings and a fixed pool. Our free opener uses rounded 0.26% gold odds, so its benchmark numbers differ slightly. No finite number of openings guarantees gold.
Published, derived and modeled are different
- Published: historical rarity ratios. StatTrak is commonly referenced as a 10% conditional chance on eligible items, not an extra rarity tier; gloves are not StatTrak eligible.
- Derived: cumulative probabilities above, assuming independent trials. Dividing a rarity’s probability evenly among items is an assumption about the pool, not a separately published per-skin figure.
- Modeled: wear, exact rare-item allocation and phases. Our calculator uses a disclosed uniform-float model over each skin’s permitted range; Valve’s actual wear distribution is not verified by that model.
A gold chance is not a profit chance. Profit requires an outcome worth more than the case plus key, with actual resale fees deducted. An expensive rare skin can raise EV while most openings still return much less than the cost.
Game edge comparison: compare rules, not marketing
There is no verified universal RTP for “CS2 case sites” or “crash.” The examples below are arithmetic illustrations, not measurements of listed operators.
| Format | What determines return | Transparent example |
|---|---|---|
| Valve weapon case | Outcome values, rarity model, case + key | EV $1.50 / cost $3 = 50% return, before fees |
| Custom case | Published per-item odds and payout values | Hidden odds → EV cannot be verified |
| Coinflip | Win probability, prize pot and rake | 50% win; 5% rake on a 2× pot → 95% return |
| Roulette | Number of equally likely slots and payout | 14 slots, 2× gross payout on 7 slots → 100%; on 6 slots → 85.71% |
| Crash | Multiplier distribution and cash-out rule | P(reaches 2×) = 49% → fixed 2× cash-out returns 98% |
Examples use gross payouts including the stake. Provably fair verification checks whether a result follows the committed algorithm; it does not prove good odds, licensing, solvency or reliable withdrawals.
Methodology & sources
Case metadata: ByMykel’s community catalog. Marketplace cache: OpenSkin’s separate CSFloat, Steam and Skinport quotes, with CSFloat listing asks used for the published report. Skin and case prices remain separate; absent values are never replaced with a proxy or zero. The scheduled service retrieves data, but market availability determines source freshness.
The report is fixed to its capture date. The free calculator reads the existing cache and displays its own dates. We have no historical series to substantiate monthly price changes, revenue totals or market-wide trends. No test purchases or actual gambling sessions were conducted for this research.