Are CS2 Case Drops Rigged? Real Drop Rates vs Official Odds
Updated · 6 min read
No evidence CS2 cases are rigged. A knife or gloves is 0.26% per case, every roll independent. See cases needed for 50/90/95% and expected value.
No evidence shows that CS2 weapon cases are rigged. Valve’s published rates are fixed at 0.26% for a knife or gloves per case (about 1 in 385), and every opening is an independent draw. At that rate it takes 267 cases for a 50% chance of at least one knife or gloves, 885 for 90% and 1,151 for 95%, so a dry streak of a few hundred cases is ordinary variance.
How do the official odds compare with what players see?
Over thousands of openings, community tracking lands close to the official table. Over a few hundred, individual results stray far from it. A Chinese Ministry of Culture law required loot-box games to publish drop likelihoods, and Valve’s local partner Perfect World released the CS:GO odds in September 2017. Those numbers matched what community trackers had estimated from large samples, and they carried over unchanged into CS2.
| Rarity | Official chance | About 1 in | Expected per 1,000 cases |
|---|---|---|---|
| Mil-Spec (blue) | 79.92% | 1.25 | ~799 |
| Restricted (purple) | 15.98% | ~6 | ~160 |
| Classified (pink) | 3.20% | ~31 | ~32 |
| Covert (red) | 0.64% | ~156 | ~6 |
| Knife or gloves (gold) | 0.26% | ~385 | ~2.6 |
These rates apply to standard weapon cases. Sticker capsules and souvenir packages use different rarity ladders. The original disclosure carried extra decimals: 79.923% for blue, 15.985% for purple, 3.197% for pink, 0.639% for red and 0.256% for gold. That rounds to the same table, and explains why some sources quote 1 in 391 instead of 1 in 385.
How does a roll work, and is there a pity timer?
Each opening is a random draw against fixed percentages, and the game keeps no count of your earlier cases. Each tier is five times rarer than the one below it, except the jump from Covert to gold, which is a 2:5 ratio. The result is rolled at the moment of unboxing.
- Tier: blue to gold, using the table above. The item pool changes from case to case; the tier odds do not.
- Item: within a tier, every item is equally likely.
- StatTrak™: eligible items have roughly a 1 in 10 chance, rolled separately.
- Wear and pattern: float is randomized within the range each skin allows.
There is no pity system. A player who has opened 384 cases without a knife has the same 0.26% chance on the 385th as on the first. Believing a hit is “due” is the gambler’s fallacy.
Why don’t a few hundred cases match the published percentages?
Percentages describe long-run averages, and short runs swing widely around them. For N openings with chance p, the typical deviation is about the square root of N × p × (1 − p).
| Sample | Expected | What randomness does |
|---|---|---|
| Blues in 100 cases | ~80 | Roughly 72 to 88 in 95% of runs |
| Knife or gloves in 100 cases | 0.26 | ~77% of runs see none |
| Knife or gloves in 385 cases | 1.0 | ~37% of runs see none |
| Knife or gloves in 1,000 cases | 2.6 | ~7% see none; ~12% see five or more |
These are binomial calculations from the 0.26% rate. A dry 300-case stretch is therefore ordinary, and even after 385 cases the chance of at least one hit is only 63.3%.
Pinning down a 0.26% event to within ±0.1 percentage points at 95% confidence takes roughly 10,000 openings. A forum thread of 200 cases cannot confirm or refute the table. To check your own history properly:
- Count every opening in one sample, including bad runs. Do not start counting after a hit.
- Multiply N by each tier’s chance to get the expected count, then compare using the deviation formula above. Two deviations away is unusual; one is routine.
- Fix the sample size first. Stopping at a good or bad moment skews the result.
- Remember you are checking five tiers at once, so one will look odd by chance.
Social feeds add survivorship bias: big pulls get shared, while thousands of quiet openings go unseen.
How many cases do you need for a knife?
No number guarantees one. The chance of at least one hit after N cases is 1 − (1 − p)^N, where p is 0.0026.
| Chance of 1+ knife or gloves | Cases at 0.26% | Cases at 0.256% | Keys alone at $2.49 (0.26%) |
|---|---|---|---|
| 50% | 267 | ~271 | ~$665 |
| 90% | 885 | ~899 | ~$2,204 |
| 95% | 1,151 | ~1,169 | ~$2,866 |
| 99% | 1,769 | ~1,797 | ~$4,405 |
Keys are bought from the in-game store for $2.49. The key column excludes the cost of the cases themselves. Even at 99%, 1 player in 100 still has no knife. Sources round slightly differently, so treat about 1,150 as the 95% answer.
The 0.26% covers every knife or glove in the pool, so any single finish is far rarer. The Kilowatt Case has 13 knives across one model (the Kukri Knife), per SteamDB. If the 0.26% splits equally, each finish is about 0.02%, roughly 1 in 5,000 cases.
Cases are no longer the only route to a knife. In its October 2025 update, Valve changed the Trade Up Contract so that 5 regular Covert items can be exchanged for one knife or gloves, and 5 StatTrak™ Covert items for one StatTrak™ knife (HLTV). That route does not use the 0.26% roll.
What is the StatTrak™ chance?
The StatTrak™ chance is about 10% per eligible drop, rolled after rarity. Perfect World’s disclosure said any grade of item has a 10% chance of dropping its StatTrak variant. Community tracking puts the empirical figure at 9.83% over 20,090 recorded openings, which supports a flat one-in-ten design (SteamDB).
| Tier | Base chance | With StatTrak™ (10%) | About 1 in |
|---|---|---|---|
| Mil-Spec | 79.92% | ~7.99% | ~13 |
| Restricted | 15.98% | ~1.60% | ~63 |
| Classified | 3.20% | 0.32% | ~313 |
| Covert | 0.64% | 0.064% | ~1,563 |
| Knife | 0.26% | ~0.026% | ~3,846 |
Gloves have no StatTrak™ version. In a case whose gold pool mixes knives and gloves, the real StatTrak™ knife chance is a little lower than the table shows.
What is a case worth against the $2.49 key?
On average a case returns less than it costs. Expected value (EV) is each tier’s chance multiplied by the average price of its items, and the Fever Case shows how skewed that payout is. With Skinport prices from 24 Sep 2026, CSRated found a Fever Case returned about 52% of its $3.02 cost with a $2.49 key.
| Tier (Fever Case) | Chance | Average item value | Adds to EV |
|---|---|---|---|
| Mil-Spec | 79.92% | $0.276 | $0.221 |
| Restricted | 15.98% | $0.889 | $0.142 |
| Classified | 3.20% | $7.106 | $0.227 |
| Covert | 0.64% | $61.54 | $0.394 |
| Knife or gloves | 0.26% | $228.33 | $0.594 |
The five rows sum to an EV of $1.578 against a $3.02 cost, an average loss of about $1.44 per opening before any selling fee (CSRated). Covert and gold outcomes, with a combined probability under 1%, supply roughly 63% of that EV. Most openings return a blue worth far less than the key.
Prices move daily and other trackers use different price sets, so the exact percentage shifts. SteamDB’s live Fever Case page showed 44 cents back per dollar at the time of writing. All of these are before fees, and the roughly 15% Steam Market fee pushes real-money returns lower.
Key takeaway: “Not rigged” means the odds are fixed and independent, not that cases are a good bet. Expected return sits below cost, and long dry streaks are normal.
Can case odds be verified, and what does provably fair add?
Aggregate odds can be checked against large community samples, but individual rolls on Valve’s servers cannot be audited. Matching a published table over thousands of openings is a statistical check, not per-roll proof.
Provably fair systems address that gap on third-party sites. At CSDROP, case openings and games use HMAC-SHA256 from a server seed (shown hashed before play), your client seed and a nonce, and results can be verified from the game history. In case battles, the client seed comes from a future EOS blockchain block ID. The method is explained on the CSDROP provably fair page.
Provably fair shows a result was not changed after it was committed. It does not make a case profitable. Each CSDROP case shows its item pool and drop rates before you open it, so read that page rather than assuming Valve’s table applies.
CSDROP is for adults only: 18+, or the age of majority in your country if higher. Set a budget before you start, treat spending as entertainment and stop when it stops being fun.
FAQ
Are CS2 cases rigged?
No evidence shows that. Valve’s published rates are fixed, and community tracking of large samples matches them. Cases do have a negative expected return, so losing on average is how they are priced, not proof of rigging.
Does a dry streak make a knife more likely?
No. Each opening is an independent draw at the same 0.26% rate, so a long streak does not make you “due”. The cumulative chance rises as you open more cases, but the next case is always 0.26%.
How many cases do I need to open to get a knife?
There is no guaranteed number. At 0.26%, about 267 cases give a 50% chance, 885 give 90% and 1,151 give 95%. Even after the “average” 385 cases, about 37% of players still have none.
Do opening times, key batches or account level change drop rates?
Valve’s published table has no modifiers for any of these, and nothing verified shows otherwise. Short-run luck explains most “systems” that seem to work.
Is the StatTrak™ chance exactly 10%?
The 10% figure comes from the same 2017 Perfect World disclosure as the tier rates. Community counts, such as 9.83% over 20,090 openings, are consistent with it. Treat it as a flat one-in-ten rule, not a figure measured to the decimal.
Are CSDROP case odds the same as Valve’s?
Do not assume so. Each case on CSDROP shows its own item pool and drop rates before you open it. To try cases first, new accounts get free cases with no deposit needed, and daily free cases unlock by level with deposit and wager conditions.
Put it to use on CSDROP
Open cases, battle other players and upgrade your skins, with provably fair results.