Wiki
CS2 Case Odds
Updated May 15, 2026 · CSGO Top Editorial Desk
Definition
CS2 case odds are the published probabilities for each rarity tier when opening a Valve case. Valve publishes the rates per tier, but the per-skin distribution within a tier is uniform. The knife/glove rate is the most-cited number; the mid-tier rates dominate the actual outcome distribution.
CS2 case odds are one of the most-discussed and most-misunderstood numbers in the skin economy. The numbers themselves are simple. The implications take some unpacking.
The published rates
Valve publishes the per-tier drop rates for cases. The standard distribution is:
- Mil-Spec (blue): 79.92%
- Restricted (purple): 15.98%
- Classified (pink): 3.20%
- Covert (red): 0.64%
- Knife / Glove: 0.26%
These numbers add to 100% — the case always drops something. The 0.26% knife rate is the most-cited number, but mathematically it’s the smallest contributor to the overall expected value of the case.
The roll mechanics
When you open a case, the game performs three independent rolls:
- Tier roll: chooses one of the five tiers using the rates above.
- Skin roll: chooses uniformly among the skins within the chosen tier.
- Float and pattern roll: independently rolls the float value (within the destination skin’s float range) and the pattern index (which controls visual variations like Case-Hardened blue percentages).
Each roll is independent. The game does not “compensate” for previous rolls — there is no streak protection, and the next case you open carries no information from previous ones.
What “0.26% knife” actually means
For an individual opening, 0.26% knife is one in roughly 384 cases. For a single session of, say, 20 cases, the odds of not pulling a knife are about 95%. The often-quoted “I opened 50 cases and got nothing” is firmly within the expected outcome distribution.
The structural reason cases are −EV per key isn’t the knife rate. It’s the combination of:
- The Steam Market fee on the eventual sale of any item (~13% effective).
- The skewed value distribution within the Mil-Spec and Restricted tiers (most of those skins trade for less than the cost of a key).
- The fact that the expected outcome — a Mil-Spec drop, 80% of the time — sells for less than the key.
Even if you pull a knife on 0.26% of cases, the remaining 99.74% of cases dominate the EV calculation.
Provable fairness, on Steam vs third parties
Valve’s case opening is not “provably fair” in the cryptographic sense. You can’t independently reconstruct the outcome of a case opening from a published seed. The community trusts Valve’s case opening because (a) the rates are published, (b) the published rates can be tested empirically across very large opening samples, and (c) Valve has an enormous reputational disincentive against fraudulent rate manipulation.
Third-party case-opening sites can implement provable fairness via published seed commits, and the better operators do. Provable fairness on a per-roll basis is meaningful for individual openings but does not constrain the pool composition — a provably-fair site can still be running pools whose advertised odds differ from real ones if the underlying database isn’t audited.
Where to verify
For Valve cases, the canonical source for case rates is Valve’s own published documentation. For third-party sites, the case detail page should publish the full pool with per-item probabilities. If a third-party site publishes case “rates” but won’t show per-item breakdowns, treat that as a meaningful credibility issue.