Nearly every box site now carries the label, so this list ranks provably fair mystery box sites on the one thing the label can settle: can you take a real result and reproduce it from published inputs? Not bonuses, not star ratings. Of the 17 platforms we track, 9 pass that test and 8 do not, and among the 11 box and card platforms specifically, 4 have been reproduced against a real outcome.
We are not an operator and do not rank boxes by their prizes. We score each site on whether its fairness claim survives a real check, using our own verifier.
What a box site owes you
Whatever a box promises, the moment that matters is the one where a number decides between the item you wanted and the one you got. On a mixed-goods site that number is the gap between the iPhone box's phone and its cable; on a card site, between the chase card and bulk. So a site owes you two separate proofs. The first is that the number was honest: committed before you clicked and unchanged after. The second is that the number maps to the item the site actually handed you, which requires a published range for every item in the box.
Plenty of sites manage the first proof and skip the second. A site that publishes a formula but no item ranges lets you confirm the roll while telling you nothing about whether that roll owed you the better prize. The sites at the top of this page publish both.
How we ranked these
What a place on this list requires
A site earns its tier from what it publishes and whether our own implementation reproduces a real outcome from it. Nothing else moves a site up or down: not the size of its catalogue, not its review score, not whether it pays an affiliate commission.
There are three tiers, and the boundary between them is precise:
- Reproduced. The published algorithm is complete enough that our implementation returns the same number the operator's own system did, on a real set of seeds. This is the only tier where the fairness claim has actually been tested rather than read.
- Published, not yet confirmed. The site documents a formula we can implement, but we do not yet have a real result to check it against, or one step of the conversion is missing.
- Not checkable. The site uses the phrase but does not publish the formula, the seeds, or both. Nothing about the claim can be tested from outside the company.
How ties are broken
Within a tier the table lists sites alphabetically, so a site's position above another in the same tier carries no judgement. The last column separates two things that are easy to conflate: reproducing a number the operator itself published, and reproducing the operator's published code without any operator number to check it against. The second is weaker, and Ripster.gg carries a further caveat, since it is operated by the company that operates CheckMyPull and its cross-check runs against that operator's own verifier.
What this ranking deliberately ignores
Prize value, deposit bonuses, shipping speed and star ratings are real considerations, and none of them moves a site between tiers. A site can be generous and unverifiable, or verifiable and disappointing, and mixing those judgements is exactly how "provably fair" got reduced to a badge. Once the tier is set, though, what a site actually sells decides which verifiable option fits you, so each entry below says what the site is as well as what it publishes.
The comparison at a glance
The registry behind this page covers 11 box and card platforms, listed here with the algorithm each one publishes. The 6 CS2 case platforms we also cover sit on the algorithms by platform page.
| # | Site | Hash it publishes | Outcome | Reproduced by us |
|---|---|---|---|---|
| 1 | HypeDrop | HMAC-SHA256, combined seed as key | Roll, scaled to 99,999,999 | Yes: stated 35,864,249 and 88.02 reproduce; unboxing 19,569,993 computed from its code |
| 2 | JemLit | HMAC-SHA512/256 | Roll, 0 to 100 | Yes, 22.0231, its own worked example |
| 3 | PackDraw | SHA-512 | Ticket, 1 to 1,000,000 | Code level, 268,760 and 122,022 |
| 4 | Ripster.gg | HMAC-SHA256 | Roll, 0 to 100 | Code level, 52.5587, not independent |
| 5 | Cases.gg | seedrandom v3.0.5 PRNG | Per-item selection | Implemented, unconfirmed |
| 6 | LuxDrop | SHA-512 | Ticket range per item | Implemented, unconfirmed |
| 7 | ArenaClub | None published | Not published | No, no seeds or hashes exist |
| 8 | Boxed.gg | HMAC-SHA512 | Ticket, 0 to 999,999 | Ticket only, item ranges unpublished |
| 9 | Pullbox.gg | HMAC-SHA512, stated only | Not published | No, code is behind a login |
| 10 | RillaBox | SHA-256 commitment only | Stated as 0 to 100 at 4 decimals | No, no box-opening formula published |
| 11 | Supabox | None published | Not published | No, no algorithm published |
Read the last column first. Four of eleven, and the four are not the four with the largest marketing budgets.
Fully verifiable: reproduced against a real result
Every site in this group reproduces in our tool, against an answer the operator states or the operator's own code, not just read. Across the full registry that group also includes the CS2 platforms: HypeDrop at 35,864,249, CSGORoll at 90,455,280, CSGOEmpire at 56,288, JemLit at 22.0231, Skin.club at 66,748, Clash.gg at 18,875. The four here sell four quite different things, so verifiability narrows the field and what you collect picks the site.
HypeDrop, if cards are the point
HypeDrop leads its homepage with cards, eight card categories plus gift cards, including sealed booster boxes, though the cards sit in front of a mostly goods shelf: on 14 October about 95 of its 855 boxes were card boxes, and the list ran from $0.16 to $70,527.34. It publishes a specific method for each game mode and exposes all three inputs you need, and both values its code states reproduce, 35,864,249 for a battle and 88.02 for an upgrade; its unboxing code gives 19,569,993 on its own example seeds. One warning matters more than the rest. Its unboxing and battle modes hash different strings, so a check run in the wrong mode returns a plausible wrong number: its own battle example gives 35,864,249 as a battle and 59,969,329 as an unboxing. The HypeDrop verifier handles the mode selection for you, and what its reviewers report about account restrictions and payouts sits in is HypeDrop legit.
JemLit, if you want to learn the check itself
JemLit sells mostly consumer goods, 334 boxes on 14 October from $0.10 to $9,950 named for their contents, plus a 15-pack card shelf, shipping to Europe, the US and parts of Asia, Africa and South America. On fairness it is the cleanest example on this page: sample verification code, per-item win ranges, and the one disclosure nobody else makes, a numbered worked example. A published set of seeds and the exact roll they produce, 22.0231, lets you confirm your understanding before trusting it on real money, which is the best single disclosure decision in the niche. Seeds live in its Fairness popup. If you want one site to learn verification on, this is it.
PackDraw, singles with a lottery on top
PackDraw runs Pokemon singles with a luxury tier above them, packs reaching $22,972.46 in September and a top-opens rail of six-figure slabs, watches and cars, paying out as a shipped item or an instant crypto withdrawal. It publishes something rarer than a description of its method: the working code its site runs, and running that code independently reproduces tickets 268,760 and 122,022. Runnable code beats a worked example, since code reproduces every case rather than one; the missing worked example is still the honest gap, because there is no known-good pair to confirm an implementation against. One practical detail: its commitment fingerprint is 128 characters rather than the usual 64, so save the whole thing. Seeds sit in your pack purchase history, and jackpot battles use a separate scheme that is implemented but not yet confirmed. See the PackDraw verifier.
Ripster.gg, if you are chasing one card
Ripster.gg is the trading card option: Pokemon, One Piece, sports cards and Disney Lorcana, boxes from $1.91 to $562.47 when last read in September, and shelves arranged by what you are chasing rather than by brand, from Holy Grails down to character Farm boxes. Its fairness page publishes the full algorithm with copy-paste verifier code and its own in-browser check, and it cross-checks against that verifier at roll 52.5587. Two caveats, stated plainly: the match is code-level rather than a captured roll from a real purchase, and the cross-check is not independent, because Ripster.gg is operated by the company that operates CheckMyPull. Its public track record is still building the way it does for any newer site; on verifiability it qualifies outright.
Published, but not yet confirmed
LuxDrop, Boxed.gg and Cases.gg give you most of what you need, but one step takes extra effort or is not yet nailed down against a real outcome.
LuxDrop publishes runnable code directly on its fairness page, a straightforward SHA-512 over serverSeed:clientSeed:nonce, mapped to a ticket range per item. Note the seed order: it is the reverse of PackDraw's, which uses the same hash and the same separator. That is the seed-order trap, and it means two sites can share every visible ingredient and produce completely unrelated results from identical inputs. Seeds are visible and there is an on-page verify button. Smaller audience than the big names, but on pure checkability it is one of the easiest to confirm once a real result is available.
Boxed.gg is the strongest TCG-focused option on documentation, covering six card games by published reviews' count, with boxes roughly $1.97 to $30.51 on its Gems currency (Deadspin) and something collectors actually use: a Marketplace where pulls are traded in, with Boxed.gg itself, toward a specific graded card rather than only shipped or sold back. On fairness it documents its inputs, reveals the server seed within 24 hours, and links code that carries the whole calculation to a ticket from 0 to 999,999. The one gap is the mapping: it does not publish which tickets belong to which item, so a check proves the number and stops one step short of the card.
Cases.gg sells fashion, tech, watches and a Pokemon shelf, naming Nike and Louis Vuitton in its own copy, with 426 boxes on 14 October from $0.10 to $14,867.84. It builds verification controls into its own fairness page, exposes seeds, and auto-rotates the server seed every 1,000 opens. Rather than hashing your inputs it feeds them to a seeded random number generator, seedrandom v3.0.5, which makes the exact library version part of the specification: a different version produces a plausible wrong number, no less random and no more right. You can check a result on-site and its inputs are reachable, though the conversion is less cleanly documented than the sites above.
Claims fairness, but you cannot check it today
RillaBox has a fairness page and describes a hashed-result scheme with a SHA-256 commitment and a result between 0 and 100 to 4 decimal places, but it publishes no formula, no combination string and no nonce for box openings, and it does not expose the seeds needed to reproduce one. Its Crash game is the exception, with runnable code on its own page, but Crash is not a box. That means the claim cannot currently be independently verified, whatever else is true about the site. This is not an accusation, it is a plain statement of what the check can and cannot reach: without visible seeds, there is nothing to recompute. If it exposes seeds later, that changes.
Pullbox.gg states that it signs your client seed, a salt and the nonce with its server seed using HMAC-SHA512, shows a SHA-256 hash of the server seed on every box page before you open, and turns the result into a slot on the box's weighted item table by rejection sampling. That is more than several sites offer. The exact byte order, the separators and the rejection rule, the details that decide the answer, are not on the page; checking a roll runs through the Verify button in your Box History. A formula an outsider cannot read is not a published formula.
ArenaClub is the clearest case on the list, because it does not claim otherwise. There are no seeds, no hashes and no verification tool. The only randomness statement we could find is a reference to a Fisher-Yates shuffle in a marketing blog post, which describes a shuffling algorithm rather than a commitment scheme, and nothing about it can be independently checked.
Supabox published no algorithm at all, and the descriptions that were accessible were partly self-contradictory, referring both to a seed commitment and to server-side random bytes. Those are two different models and only one of them is checkable. As of October 2026 the supabox.com address forwards to cravin.com, and the page it lands on promotes skill-based games rather than boxes.
The same logic rules out any site that publishes only pull rates with no seed-commit scheme. An odds table is useful, and a fairness page with no visible seeds is a badge, not a proof.
Where to find your seeds on each site
Knowing a site is verifiable is only useful if you can find the three values. The panel has a different name on every platform, and the revealed server seed almost never appears until you do something to trigger it.
| Site | Where the commitment sits | How the real seed is revealed |
|---|---|---|
| Ripster.gg | Provably Fair modal, before you roll | Same modal after the roll. Changing your client seed in Account Settings reveals it for all past rolls |
| PackDraw | Your account profile, before purchase | Exposed once the purchase completes. Per-open seeds are in your pack purchase history |
| HypeDrop | Shown before the unbox, with the client seed editable | Shown after. Battle seeds are in the Provably Fair modal on the Battle Duel page |
| JemLit | The Fairness popup | The same popup, after rotation |
| Boxed.gg | Shown hashed before you open | Revealed every 24 hours. Per-item ranges are under More Info on the drop table |
| Cases.gg | Its own fairness page | The server seed rotates automatically every 1,000 opens |
The step almost everyone misses
Rotation. On most of these sites the real server seed stays hidden until you retire it, which means saving the hashed value before you play and rotating afterward are both part of the same check. Skip the first and you have nothing to compare against. Skip the second and there is nothing to compare.
Boxed.gg and Cases.gg are the exceptions, rotating on a schedule rather than on demand: every 24 hours and every 1,000 opens. On those two you wait rather than act, but the sequence is otherwise identical.
How to check any of these yourself
Wherever a site exposes seeds, the check is the same, and you can do it right here. The verifier below runs entirely on your own device.
- Before you open, save the hashed server seed. This is the commitment.
- Note your client seed and the play count for the pull you care about.
- Rotate your seed so the site reveals the real server seed.
- Paste the values into the verifier. It recomputes the outcome and confirms the revealed seed matches the hash you saved.
Step 1Where did you play?
Using Ripster box / upgrade: HMAC-SHA256 over clientSeed:noncedetails
HMAC-SHA256 keyed with the server seed over clientSeed:nonce. The first 8 hex characters become a 32-bit integer, divided by 2^32 and multiplied by 100 for a roll in [0, 100). Items are sorted by item ID and selected by cumulative probability: the first item whose cumulative probability is at least the roll wins.
Where to find your numbers on Ripster.gg:
- Before rolling, copy the Server Seed Hash from the Provably Fair modal (also recorded per roll in Account History).
- After the roll, open the Provably Fair modal to copy the Server Seed, Client Seed and Nonce.
- Change your Client Seed anytime in Account Settings; doing so reveals the current Server Seed so all past rolls become verifiable.
Source: Ripster.gg's own fairness page
Step 2Paste your numbers
Runs entirely on your device via your browser's built-in cryptography. Don't take our word for it: open DevTools → Network, click the button, and watch: zero requests.
Runs entirely in your browser. Nothing you paste is sent anywhere.
Pick the right platform before you paste. The outcome ranges above are not interchangeable, and a verifier set to the wrong site returns a plausible wrong number. If you are new to what these inputs are, the what is provably fair guide walks through the whole flow, and what is a server seed covers the single most important input in detail.
What this list does not tell you
The bottom line
If your priority is being able to check what you get, 4 of the 11 box and card platforms we cover clear the bar today: JemLit, PackDraw, HypeDrop and Ripster.gg all publish enough that our implementation reproduces a real outcome from their own inputs. LuxDrop, Boxed.gg and Cases.gg publish a formula we can implement but are not yet confirmed end to end. RillaBox, Pullbox.gg, ArenaClub and Supabox cannot be checked from outside at all, whatever their fairness pages say.
Whichever you use, do not take the label on trust. Rotate your seeds, paste a real result into the verifier, and watch it check out for yourself. If any part of this has stopped being fun, take a break and get support.