The gap between published odds and verifiable odds is the difference between being told a rate and being able to check one. A sealed booster publishes pull rates you can never test on an individual pack. A provably fair box publishes ticket ranges you can recompute for one specific opening, in about a second. Both print odds; only one lets you audit a result.
What a sealed pack promises
Open a sealed booster and the manufacturer has told you, in advance, roughly how often each rarity appears. Those pull rates are real information. They are also, for any individual pack, entirely uncheckable.
There is no value committed to your specific pack that you can inspect afterwards. There is no fingerprint published before you opened it. The cards were placed during manufacturing, and once the wrapper is off, the only evidence you have is the cards themselves. You cannot prove your pack was drawn honestly, and you cannot prove it was not.
That is not an accusation. It is a structural fact about physical products. The promise a sealed pack makes is a statistical one: across enough packs, the rates should hold.
Which means the only honest way to test it is statistically. Open a meaningful number, record what you actually pulled, and compare that against what the published rates predict.
Why "a meaningful number" is bigger than it feels
The trouble is that "a meaningful number" is far larger than most people assume, and that gap is where almost every rigging accusation comes from. Take an item listed at 0.5%, one in two hundred:
| Openings | Chance of seeing zero of them |
|---|---|
| 50 | 77.8% |
| 100 | 60.6% |
| 200 | 36.7% |
Open a hundred and miss it entirely and you have hit the most likely outcome. Nothing has gone wrong. To reach even a 95% chance of pulling one at least once, you need roughly 598 openings.
The numbers move fast with rarity. At 1% the same 95% confidence arrives at about 298 openings; at 0.1% it takes about 2,994. So a run of 50 or 100 that produced nothing good is not evidence of anything at all, and a run that produced two chase items is not evidence either.
That comparison is what the variance checker does: it takes your actual pull history and tells you whether it sits inside the range those published rates would ordinarily produce, or genuinely outside it. It works on any history you have, physical or digital.
Anyone telling you a single unlucky pack proves manipulation is overreading the evidence by a wide margin. A single pack cannot prove anything either way. That is the whole point.
What a provably fair box promises
A provably fair site makes a different and much narrower promise, and it keeps it in a way you can check on one opening.
Before you open anything, the site publishes a fingerprint of a secret value. After you rotate seeds, it reveals that value. You can hash the revealed value and confirm it matches the fingerprint published earlier, which proves the site was locked in before you played. Then you can feed the revealed value, your own seed and a counter into the published formula and recompute the exact ticket your opening produced.
Every item holds a published range of tickets. Yours landed in one of them. If the recomputed ticket falls in the range for the item you actually received, the result is confirmed: it came from values fixed before you clicked, and nothing was adjusted afterwards.
What the ticket space actually looks like
The ticket space is what makes this concrete rather than abstract. PackDraw normalises every pack to a ticket from 1 to 1,000,000, so an item listed at 0.5% holds exactly 5,000 of them. Clash.gg uses 0 to 99,999, so a 1% item holds 1,000 tickets and a 0.1% item holds 100. Skin.club runs 1 to 100,000. Different ceilings, same idea: the published chance is not a claim about the long run, it is a literal count of the numbers that win.
That is a per-outcome promise. One opening, checked, settled. No sample size required, no 598 openings, no waiting.
The comparison
| Sealed pack | Provably fair box | |
|---|---|---|
| Odds published | Yes | Yes |
| Can you check one opening? | No | Yes |
| Can you check the long run? | Yes, with enough data | Yes |
| What is committed in advance? | Nothing you can inspect | A fingerprint of the secret value |
| Evidence after the fact | The cards | Seeds, ticket, published ranges |
The row that matters is the second one. It is the entire difference, and it is invisible if you only compare the published odds.
How to check it yourself
If you have a result from a provably fair site, you do not have to take any of this on trust. Take the revealed server seed, your client seed and the nonce, and recompute the outcome here. It runs on your own device, so nothing you paste reaches us or the operator.
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.
That is the whole difference in practice. A sealed pack offers nothing to paste in; there is no committed value and no formula, so the box above has nothing to work with. A provably fair result gives you three values and a published algorithm, and the answer arrives in about a second.
The third question neither answers
Both of the above tell you whether the stated odds were honoured. Neither tells you whether the stated odds are any good.
That is a separate question, with separate evidence: what an item is actually worth, how the rates compare across sites, whether the thing you are chasing is realistically reachable. A verifier is silent on all of it. So is a pull-rate table. Treating a passing fairness check as a general endorsement is the most common mistake people make with these systems, and it is why we keep what provably fair does not prove as a standing page rather than a footnote.
Which check applies to you
Work out which promise you were given, and the right check follows immediately.
No committed value, no revealed seed, no published ticket ranges? You have a statistical promise. One opening proves nothing. Collect a real sample and run it through the variance checker.
A fingerprint before, a revealed seed after, published ranges per item? You have a per-outcome promise. Take the seeds and recompute the result with the verifier, and you will know about that specific opening in seconds.
A site that claims provably fair but gives you no seeds to check? That is the case worth paying attention to. The claim is doing work the evidence does not support, and spotting the difference is a skill worth having.
You can try the first kind for yourself without spending anything: the pack simulators run real set lists and published pull rates, so you can see how much a run of openings swings around before drawing conclusions from your own.
The bottom line
Published odds and verifiable odds are two different promises. A sealed pack gives you a rate that only means anything across hundreds of openings: at 0.5% you need roughly 598 to reach 95% confidence of a single hit. A provably fair box gives you a committed value, a revealed seed and a published ticket range, so one opening settles one question in seconds.
Work out which promise you were handed, then run the matching check. If there are seeds to paste, use the verifier and get an answer about that specific result. If there are not, collect a real sample and run it through the variance checker instead of reading a bad night as proof of anything.