CS2 case odds and drop rates explained

Every question about CS2 case opening eventually reduces to one number: the probability of the drop you are hoping for. This guide lays out the real numbers — Valve's officially disclosed drop rates, the knife odds everyone actually asks about, how third-party case sites structure their odds differently, and how to read any odds table well enough to calculate what a case is really worth before you open it.

A short note on where the official numbers come from, because it matters for trusting them: Valve never volunteered its drop rates. They became public in 2017, when China's regulator required loot box probability disclosure and Valve published the figures through Perfect World, its Chinese distribution partner. The same rates have applied unchanged from CS:GO through CS2, and community tracking of millions of recorded openings keeps matching the published table — which is the ongoing statistical check on a system that offers no per-roll verification.

One framing rule before the tables: every number in this guide describes a negative-expected-value activity. No official case and no honest third-party case returns more than it costs on average. Odds knowledge will not make case opening profitable; what it does is tell you exactly what you are paying for, and it makes inflated marketing claims immediately recognizable.

Every official CS2 weapon case draws from the same five-tier probability table. The case's item pool changes; the tier odds never do. These are Valve's disclosed rates:

Two further rolls happen after the tier is decided. First, every eligible drop has a 10% chance of being the StatTrak version, which tracks kills and typically raises market value. Second, the item receives a float value between 0.00 and 1.00 that maps to the five wear grades from Factory New to Battle-Scarred — and wear can move a skin's price by several multiples.

A common confusion worth clearing up: the CS2 rarity ladder also contains Consumer (white) and Industrial (light blue) grades below Mil-Spec. Those tiers do not drop from weapon cases at all — they come from collections and weekly drops. If an odds table for a "case" shows Consumer-grade items, you are looking at a third-party custom case, not an official Valve one.

The 0.26% special-item rate is the single most-quoted number in case opening, so it deserves the practical translation. One in 385 means that if you opened one case per day, you would wait a median of about nine months for a knife or gloves — and that is a median, not a guarantee. At roughly $2.50 per key plus the case price, a naive "open until knife" strategy prices a single gold drop at over $1,000 in expected spend, frequently for a knife worth a fraction of that.

Two properties of the system make intuition fail here. There is no pity timer: unlike many modern loot box systems, CS2 case odds do not improve with consecutive misses — the 386th case after 385 blanks is still 0.26%. And every opening is independent: a gold drop yesterday neither raises nor lowers today's odds. Feeling "due" for a knife is the gambler's fallacy operating exactly as designed.

The same math explains why the gold tier is not automatically the value tier. A 0.26% chance at a knife pool averaging $200 contributes about $0.52 of expected value per opening — often less than the Mil-Spec tier contributes through sheer frequency. Which leads directly to the next section: item prices matter more than tier colors.

Third-party case opening platforms run the same core mechanic — pay a fixed price, receive one random item — but the odds structure differs from Valve's in three ways that change how you evaluate them.

First, odds are published per case, before you open. Valve disclosed one table once, under regulatory pressure; serious third-party sites display the exact probability of every individual item in every case on the case page itself. That is a genuine transparency improvement — and it is also the minimum bar. A third-party site that does not show per-item odds has less disclosure than Valve, not more.

Second, the effective return is a design choice per platform. Because operators build their own cases, each case's expected value — its RTP, return to player — is set by the item pool and the odds. The only platform in CaseRadar's coverage with a well-documented independently measured figure is DatDrop (9.3/10), at approximately 80.6% across its catalog, against the roughly 50–70% that independent analyses measure for official Valve cases. The difference between 100% and the RTP is the house edge — the operator's structural margin on every opening.

Third, the serious platforms make individual results verifiable. Valve's roll happens server-side with no audit path; provably fair platforms commit to a hashed server seed before the roll and reveal it after, so you can recompute any result yourself. The full mechanism is in the provably fair guide. Of the 6 platforms CaseRadar tracks, 5 run documented provably fair systems — and the one that does not (Farmskins) also publishes no per-case odds, which is the combination to avoid.

One heuristic worth keeping: measured third-party RTPs cluster around 75–85%. A platform advertising returns above 95% — or implying profitability — is advertising something no honest case economics produce. Treat that claim as a fraud signal, not a bargain.

Every legitimate third-party case page shows the same underlying data: a list of items, each with a market value and a drop probability. Reading it takes three steps and no advanced math:

The step people skip is checking current market prices. An odds table is static; skin prices move daily. A case whose headline item has crashed in price can quietly drop from 80% RTP to 60% without a single odds change — which is why the item's market price matters more than its rarity tier. A 3.20% Classified drop worth $40 contributes more expected value than a 0.64% Covert drop worth $15, every time. Tier colors describe scarcity within the case's pool; only price describes what the drop is worth to you.

As an investment, no — and this is arithmetic, not opinion. Official cases return roughly 50–70% of spend on average; measured third-party catalogs run around 75–85% at best. Every case, everywhere, prices in an operator margin. Opening cases with money you need back is strictly worse than buying the skin you want directly off a marketplace, in every scenario except the lucky one you cannot buy on purpose.

As entertainment, the honest framing is that you are paying for variance: the average opening loses 20–50% of its cost, and what you get for that is the small chance at an outsized drop plus the moment of the reveal. Third-party platforms offer a measurably better version of that deal than official cases — higher measured returns, no key cost, entry under $1, per-case odds published, and verifiable rolls on provably fair platforms — but a better version of a losing bet is still a losing bet.

If you do open cases, the two rules that follow from everything above: only spend what you can afford to lose entirely, and only play where you can check both the odds and the results — published per-item odds, and a documented provably fair system. Those criteria are exactly what CaseRadar's scoring methodology weights heaviest, and every platform review documents both.

If you came here deciding where to open cases, the ranked comparison scores every tracked platform on fairness, odds transparency, and measured value — the same criteria this guide is built on. The two highest-scored platforms: