How case site odds actually work
A case is a weighted item pool. Each item has a published probability and a value, and the sum of value times probability is the case's expected return. The house edge is the gap between that figure and the case price — typically 8 to 20%. The item shown in the marketing usually contributes almost nothing to it.
WagerSave Editorial·Updated ·7 min read·VerticalsCalculating what a case is worth
Pool EV = Σ (item value × item probability) House edge = (Case price − Pool EV) ÷ Case price
A worked case priced at 10:
House edge = (10 − 7.69) ÷ 10 = 23.1%
Every 1,000 spent on this case returns 769 in expected item value. It costs 231.
Look at the knife line. It's the entire marketing of the case, and it contributes 0.80 of the 7.69 — about a tenth. The bulk of the return sits in the bottom two rows, which are items you already have and don't want.
Odds are per open, not cumulative
The most persistent misreading in the vertical.
A 0.04% knife chance does not mean you'll hit one in 2,500 opens. Each open is independent, and 2,500 opens is the point at which the expected number of knives reaches one — not where you're guaranteed one.
P(at least one knife in n opens) = 1 − (1 − 0.0004)^n
At 2,500 opens — the "one in 2,500" point — you have a 63% chance, not a certainty. More than a third of players who open 2,500 cases hit nothing.
And 25,000 spent chasing a 2,000 knife is a bad trade at any probability. The arithmetic never favours it.
There is no pity timer, no due-a-win state, no accumulated progress. Case 4,000 has exactly the same odds as case 1.
Where sites can quietly widen the edge
The published odds are usually accurate. The published values are where the real variance in fairness sits.
Inflated item valuations. Sites price items themselves. If a skin worth 35 on the market is valued at 40 in the case pool, the calculated EV overstates what you actually get. A consistent 10% inflation on a case with a nominal 15% edge makes the real edge closer to 23%.
Float and wear not specified. A skin's condition dramatically affects its value — Factory New versus Battle-Scarred can be a 5x difference. If the pool lists an item without specifying wear, the site can deliver the worst tier while advertising the value of the best.
Buy-back rates. Many sites offer instant sell at a discount to the listed value. Winning a 40 item and selling it back at 32 means the effective pool EV was 20% lower than published.
Withdrawal restrictions. If items can't be withdrawn below a value threshold, low-tier wins are effectively locked into further play. Their contribution to the pool EV is real on paper and much less real in practice.
Check all four before trusting a calculated EV. The odds table alone doesn't tell you what a case costs.
Provably fair verifies the roll, not the pricing
Case sites publish a hashed server seed before the roll, combine it with a client seed you control and an incrementing nonce, and let you verify afterwards that the result matched.
This proves the outcome wasn't altered after you paid. It's a real guarantee and worth having.
It proves nothing about whether the pool EV is reasonable, whether the item valuations are accurate, or whether the odds are generous. A case with a 40% house edge and inflated valuations can be flawlessly provably fair.
Verify the roll. Then calculate the EV separately, because those are two independent questions and only the second one tells you what you're paying.
How odds interact with cashback
Cashback is a share of the house edge, so a wider edge produces a larger rebate:
Cashback = Turnover × House edge × Rate
Over 1,000 of opening at a 10% rate:
The 23% case pays the most cashback and leaves you worst off. This is the same trap as high-edge slots, amplified — case edges span a much wider range than slot RTPs do.
Calculate the EV, then take the cashback. Never the other way round.
The comparison nobody makes
If you want a specific skin, compare two routes:
Buy it directly. Market price, one transaction, guaranteed.
Open cases for it. Expected cost = item value ÷ probability, plus the value of everything else you receive along the way at whatever rate you can liquidate it.
For the knife above: 2,000 ÷ 0.0004 = 5,000,000 of opening in expectation, against a 2,000 purchase. You do recover roughly 77% of that in other items, so the real net cost is closer to 1,150,000 — still three orders of magnitude worse than buying it.
Case opening is entertainment priced as a lottery. That's a legitimate thing to buy if it's what you want. It is not an acquisition strategy, and no cashback rate changes that.
Frequently asked questions
How are CS2 case site odds calculated?
Each item in the pool has a fixed probability. The case's expected value is the sum of each item's value multiplied by its probability, and the house edge is the gap between that and the case price.
Are case site odds rigged?
Reputable sites publish accurate odds and provably fair verification. The more common issue isn't rigged rolls but inflated item valuations, unspecified float, and discounted buy-back rates, all of which widen the real edge without touching the published probabilities.
Does opening more cases improve my chances?
Not per open. Each roll is independent with identical odds. Cumulative probability of hitting a rare item rises with volume, but so does total cost, and the cost rises faster.
What is a typical case site house edge?
Usually 8–20%, considerably wider than casino games. Once inflated valuations and buy-back discounts are accounted for, the effective figure is often higher.
18+. Case opening carries one of the widest house edges in online gambling. If gambling stops being entertainment, find support.