CS2REF TOOL
CS2 Case Opening Simulator
Open cases without spending anything, then watch what happens to the return as the number of openings grows. The point is not the profit — it is the convergence: a short run can end up well ahead, and a long run cannot.
Last reviewed: September 18, 2026 (2026-09-18)
A modelled 90% return, the shape third-party case sites are commonly assumed to run. Each platform sets its own odds and prices.
Spent
$2500.00
Received
$3026.03
Profit
+$526.03
Your RTP
121.0%
Openings in profit
55 / 1000
Best drop
Knife / Gloves ×206.5
Read this before drawing a conclusion: 121.0% is the result of one run of 1,000 simulated openings, not a property of any real platform. Over a small number of openings the realised figure swings far from the modelled 90.0% — that swing is the point of the simulator, and it is why one person's screenshot of a big win is not evidence about the odds.
| Rarity tier | Valve probability | Modelled payout | Hits in this run | Expected hits |
|---|---|---|---|---|
| Mil-Spec | 79.92% | ×0.11 of cost | 787 | 799.2 |
| Restricted | 15.98% | ×0.56 of cost | 158 | 159.8 |
| Classified | 3.20% | ×2.35 of cost | 45 | 32.0 |
| Covert | 0.64% | ×16.89 of cost | 6 | 6.4 |
| Knife / Gloves | 0.26% | ×206.47 of cost | 4 | 2.6 |
How the simulation works
- 1Take Valve's published per-rarity drop rates — Mil-Spec, Restricted, Classified, Covert, and Knife or Gloves — which are the only distribution with a primary public source.
- 2Assign a payout shape to those tiers so the rare tier carries most of the value, matching how real case returns are distributed.
- 3Scale that shape by a single factor so the long-run average equals the modelled return you selected on the control above.
- 4Run your chosen number of openings with a seeded generator, so the same seed always reproduces the same run and you can re-run it yourself.
- 5Compare the realised return of that run with the modelled return: the gap is variance, and it shrinks as the number of openings grows.
Why the number swings so much
Almost all of the expected value sits in one tier that lands rarely. That structure has a specific consequence: the median outcome of a short session is well below the average, because the average is being held up by an outcome most sessions never reach. A run of a few dozen openings is far more likely to look bad than good, and a single rare drop can make a whole run look excellent.
This is why a screenshot is not evidence. To learn anything about the odds behind a platform you need the published probabilities and a large number of openings — not the best result anyone has ever seen.
What this simulator does not tell you
- It does not describe any specific platform. Every platform sets its own item pool, its own prices and its own odds.
- It does not know the real cash-out value of the item you receive. Only a marketplace price at the moment you sell gives you that.
- It does not prove anything about a platform's randomness. A provably-fair scheme is a separate check, and it says nothing about the payout percentage.
- It does not predict your session. It shows the shape of the distribution your session is drawn from.