CS2 Case Site House Edge: Which Case Opening Site Gives the Most Back?
Some case opening sites publish the odds and the value of every item in every case. That is enough to work out what a case gives back on average, and how much the site keeps.
We did it for 194 cases on 5 sites. On the sites' own numbers the median house edge runs from 6.0% on OpenStars (ex BountyStars) to 11.1% on G4Skins. Priced at Skinport instead, the order changes.
5 case sites194 cases checked6.0% lowest median edge18+ only
Lowest house edge first
Case sites ranked by median house edge
House edge = 1 − (expected value of one opening ÷ case price), from each site's own odds and item values. Lower is better for you. The median is the middle case of those we checked on the site.
Checked but not ranked (the published numbers fail a plausibility check, see below)
–DROP.SKIN
−2.3%return 102.3%
−73.3% to 40.8%
+34%2608 items
–
65of 65
–PandaSkins
28.6%return 71.4%
11.1% to 59.0%
−10%1080 items
–
30of 189
"Item values vs Skinport": each site's value for an item against what the same item sells for on Skinport (its median sale, never above the cheapest listing), median over the items we could match. A plus means the site values its skins above Skinport, which makes its cases look better than they are. "Return at Skinport prices": the median case's expected value with its drops priced at Skinport instead (cases where at least 95% of the odds could be matched). A case that would return more than 100% at Skinport prices is shown as "not checked" and left out of the median until its items are checked by hand: such a figure almost always comes from a thin or stale price.
At market prices, Skin.Club gives the most backMedian case return per site: on the site’s own item values (left) and with the same drops priced at Skinport (right). Lines that cross = the order changes.
Change at Skinport prices
Skin.Club−15.8 ptsitem values +20% vs Skinport
OpenStars−25.6 ptsitem values +36% vs Skinport
G4Skins−26.1 ptsitem values +39% vs Skinport
Not shown: DROP.SKIN and PandaSkins, whose published numbers fail the plausibility check (listed but not ranked, see below).
At Skinport prices the order is Skin.Club 76.7%, OpenStars (ex BountyStars) 68.4%, G4Skins 62.8%.
OpenStars (ex BountyStars): median house edge 6.0%
29 cases checked. Sample: systematic sample: 30 paid cases evenly spaced through the operator's case list (no opening counts are published). Average edge 6.8%. None of the cases we checked returns more than it costs. Prices in USD.
The site's own case data also states a house edge for every case (6%, 8%, 10%, 12%): our figure matches it within 0.1 points for 29 of 29 cases.
40 cases checked. Sample: systematic sample: 40 cases evenly spaced through the operator's case sitemap (opening counts are only on case pages, so "most popular" can't be chosen without fetching every case). Average edge 7.7%, 8.5% weighted by how often each case has been opened. None of the cases we checked returns more than it costs. Prices in USD.
30 cases checked. Sample: the 30 most-opened paid cases by the operator's own "opened" counter. Average edge 11.7%, 11.4% weighted by how often each case has been opened. None of the cases we checked returns more than it costs. Prices in USD.
65 cases checked. Sample: every paid case in the operator's case list. Average edge −3.6%. 36 of the cases we checked return more than they cost at the site's own values. Prices in USD.
Not ranked: 36 of 65 checked cases return more than they cost at the site's own values.
30 cases checked. Sample: systematic sample: 30 paid cases evenly spaced through the operator's home-page case grid (no opening counts are published). Average edge 27.5%. None of the cases we checked returns more than it costs. Prices in EUR.
Market check compared with the operator's own USD figure shown next to each EUR value.
Not ranked: 79.1% of its item values are more than 25% away from its own typical markup over Skinport sale values, so its values are not a consistent measure of what the drops are worth.
Only case sites whose public case pages carry, in the page HTML the site serves, the odds of every item, the value of every item and the case price. robots.txt allowed every page read; no login, no cookies, no CAPTCHA, no bot-check bypassing, at least the robots.txt Crawl-delay (5 s default) between requests.
Expected value (EV) of one opening = sum over every item of the item's stated odds × the item's stated value. All numbers are the operator's own, as shown on the page when fetched.
Return = EV ÷ case price. House edge = 1 − return. Free, deposit-only and event-currency cases are left out.
A case counts only if its stated odds add up to 100% (± 0.5 pp). Cases that fail are listed but left out of the operator summary.
Operator values are what the site credits for the skin (sell-back or listed value), not what the skin sells for elsewhere. The market check values the same items at what they sell for on Skinport (USD): the median Skinport sale over 30 days (at least 3 sales; otherwise over 90 days), never above the cheapest Skinport listing minus 1 cent, the same rule as our case odds pages and trade-up calculator. Items with fewer than 3 Skinport sales in 90 days and Doppler-type phases (Skinport prices each phase separately) are left out of the check and counted. An item with fewer than 3 sales in 30 days is used only when its 90-day value is within 2 times Skinport's own suggested price either way; otherwise a handful of sales would re-price the whole case, so it is left out and counted too.
A case is re-priced at Skinport only when at least 95% of its odds could be matched. A case that would return more than 100% at Skinport prices is shown as "not checked" and left out of every median until we have checked its items by hand: such a figure almost always means a thin or stale price, not free money.
Which cases: OpenStars (ex BountyStars): systematic sample: 30 paid cases evenly spaced through the operator's case list (no opening counts are published); Skin.Club: systematic sample: 40 cases evenly spaced through the operator's case sitemap (opening counts are only on case pages, so "most popular" can't be chosen without fetching every case); G4Skins: the 30 most-opened paid cases by the operator's own "opened" counter; DROP.SKIN: every paid case in the operator's case list; PandaSkins: systematic sample: 30 paid cases evenly spaced through the operator's home-page case grid (no opening counts are published).
Plausibility check: a site is listed but not ranked if more than 10% of its checked cases return over 100% at its own values, or if more than 40% of its item values (items with at least 5 Skinport listings) are over 25% away from the site's own typical ratio to Skinport sale values, i.e. it values items inconsistently. A site that values every item at a steady markup over the market is ranked, and that markup shows in the "Item values vs Skinport" and "Return at Skinport prices" columns.
Caveats
The sites set their own item values. We compared 7013 item values with what the same items sell for on Skinport: the median Skinport sale, never above the cheapest listing (listings from 18 November 2026, sales from the 18 November 2026 snapshot). Median difference: OpenStars (ex BountyStars) +36%, Skin.Club +20%, G4Skins +39%, DROP.SKIN +34%, PandaSkins −10%. A site that values items above the market shows a lower house edge than you get if you sell the skin elsewhere.
Skinport prices are not cash. Skinport values here are recent sale medians capped at the cheapest listing, before Skinport's selling fee; rare high-value items sell rarely, and items with fewer than 3 sales in 90 days are left out. Treat the "at Skinport prices" column as an estimate.
A sample, not every case. Several sites run hundreds of cases. The rule for each site is above; cases are never chosen by result.
Odds and values change. Sites re-price items and re-balance cases often. Every figure is dated; check the case page before you open.
Averages over many openings. One opening usually returns less than the case costs, often far less. The expected value only holds over a very large number of openings.
Sites we couldn't check. Many case sites load their odds with JavaScript only, block automated visitors, or don't publish per-item odds and values. We don't log in, solve CAPTCHAs or get round bot checks, so those sites aren't here.
Case opening is gambling, and the expected result is a loss. 18+ only. For Steam's own weapon cases, see CS2 case odds and expected value.
Case-opening sites sell their own cases, not Valve’s, with their own odds. These three lead our case-opening list; each review says which odds and fairness proof the site publishes.