Every platform here sells randomised packs. What separates them is whether you can find out, before spending, what the odds actually are — and whether that disclosure is precise enough to compute with.
This grades 7 platforms on exactly that. Each grade is based on evidence we gathered ourselves from live purchase screens, and each says what would have to change to improve it.
Nobody has an A
Nobody has an A, and that is not a slight on the platforms — an A requires us to have audited published odds against observed outcomes, and we have not completed that work on any platform yet. The grade measures what we can verify, not what we suspect.
Per-card pull odds across all 364 cards in the pool, plus five tier probabilities with value bands, an average pull value and the buyback rate — all on the purchase screen
Why this grade
The most granular disclosure we have found anywhere in this category, and it is not close. Individual card probabilities down to 0.0032% are published before you spend.
To improve it
An audit of published odds against observed pulls. Their robots.txt blocks our automated access, which we respect, so this has to be done by hand.
Four tier probabilities with value bands, the platform's own expected value in dollars, a 'big win chance', and the buyback rate — per machine, before purchase
Why this grade
Publishing an expected value and a buyback rate on the same panel is unusual and genuinely useful: it lets anyone compute the realizable return without guessing. Draws settle through a verifiable random function with published tier ranges.
To improve it
Our own audit of published odds against their public pull feed, which their documented API would support.
Five tier probabilities with value bands and an average value, on every Claw
Why this grade
Clean disclosure that sums to exactly 100% on all four packs we captured — tidier rounding than most. Marked down from anything higher because there is no per-pull feed to check outcomes against.
To improve it
A published swap rate alongside the odds. Without it a buyer cannot compute what they actually realise, only what the pack nominally contains.
Five value bands with probabilities and an average item value, per pack, calculated in real time
Why this grade
The only platform we have found whose published average value is BELOW the pack price on every pack sampled — it tells you the pack is worth less than it costs rather than advertising a surplus it later takes back. Pack opens settle on Polygon.
To improve it
Publishing the instant-buyback rate alongside the odds. It is the number that decides the real return and it is not shown where the decision is made.
Evidence captured 2026-08-05
Renaiss
C
What they publish
Tier probabilities with very wide value bands; the top tier shown as 'less than 1%' rather than a figure
Why this grade
A 67% chance of a card worth $35–$200 on a $150 pack is disclosure in form only. That band spans a 77% loss to a 33% gain and the buyer cannot tell which end is likely, which is the information that actually matters.
To improve it
Narrower bands and an exact top-tier probability. The structure is already there; the resolution is not.
Evidence captured 2026-08-04
Phygitals
Pending
What they publish
Odds published per pack; not yet captured first-hand by us
Why this grade
Ungraded pending a response from the platform. We observed a free demo returning a top-tier pull three times consecutively on a $50 pack — an outcome with roughly one-in-a-million odds at a typical 1% top-tier rate — which suggests the demo does not run on live odds. We have put that to them and will grade once they reply or the deadline passes.
To improve it
A response on the demo question, and our own capture of the live pack odds.
Evidence captured 2026-08-04
gacha.game
Pending
What they publish
Guaranteed-value pack with stated card values
Why this grade
Ungraded pending a response. On their $50 guaranteed pack, both possible cards appear to trade below $50 according to the PSA sales history reachable from their own listings — which would mean the guarantee does not hold. We have asked them to explain the calculation.
To improve it
A response on how the guarantee and the stated 110% return rate are calculated.
Evidence captured 2026-08-05
The rubric
Grading rubric for odds transparency
Grade
Meaning
A
Odds published, independently verifiable, and confirmed consistent with observed outcomes over time.
B
Odds published in a usable form, but we have not independently verified outcomes against them.
C
Odds published in a form too coarse to compute with — bands so wide the buyer learns little.
D
Odds not published at all.
F
Published odds contradicted by evidence.
What this index does and does not measure
It measures disclosure quality — whether the numbers you need are published, and in a form you can use. That is deliberately narrow, and worth being clear about.
·A high grade is not a recommendation. Every platform here sells packs that return less than they cost. MNSTR grades well because it publishes more than anyone else, not because its packs are a better deal.
·A low grade is not an accusation. Renaiss publishes odds; they are simply too coarse to compute with. That is a design failure, not evidence of anything worse.
·Pending means pending. Two platforms are ungraded because we have put specific questions to them and are waiting for a reply before publishing a judgement. Their answers will appear here.
If you operate one of these platforms and think a grade is wrong, the fastest route to changing it is evidence: a link to the disclosure we missed, or a change you have shipped. We re-check monthly, we publish what changed and when, and we will say plainly when a platform improved rather than quietly editing the letter.
Every grade above names what would raise it. None of them are unreachable.
Updated 2026-08-05 · re-checked monthly · grades assigned only from evidence we gathered ourselves