A procedure, not a verdict
The round you can recompute, and what that never touches
Twenty-one of the hundred operators in our library are recorded as publishing a seed-checkable procedure, and eight of the eleven in our table are among them. What that procedure lets a player do is recompute one completed round from values the operator published. What it does not do is establish the payout row, the long-run return, or that money leaves the account.
What is being verified
A seed-checkable round, usually advertised as "provably fair", is a procedure rather than a claim. It works the same way wherever it appears.
Before a round, the operator publishes a hash of a server seed — a commitment to a value it cannot then change without the hash changing. The player supplies or is given a client seed. After the round, the operator reveals the server seed, and anyone can combine the two values, run the published function, and see whether it produces the result they were shown.
If it does, the result was fixed before the bet was placed. If it does not, something was substituted.
That is the entire content of the check, and it is genuinely useful: it removes one specific way a game could be rigged, which is choosing the outcome after seeing the stake.
Who publishes one here
Our data records the procedure as a tag on the operator, not as something anyone at this site ran.
Eight of the eleven carry it: DuckDice, Metaspins, Rainbet, Shuffle, Wild.io, Wolf.bet, Bitsler and Flush.
Three do not: Bitcasino.io, Rocketpot, and Vave — the paid placement, whose record carries no tags at all.
Across the whole library the tag is on 21 of the 100 records, and 12 records carry both this tag and the in-house originals tag. That overlap is the population where an operator both builds its own board and publishes a way of re-checking a round on it, and it is discussed further on the page about in-house games and outside studios.
Three things the check does not reach
This is the part that gets lost, and it is the reason this page exists rather than a paragraph elsewhere.
It does not establish the payout row. The multipliers in the row are published by whoever built the board, and the seed procedure verifies which bin was reached — not what that bin was worth. A verified round in a row of numbers nobody can audit is a verified round in an unaudited price list. The nature of that list is set out on the page about the payout row.
It does not establish a rate. Recomputing one round says nothing about how often each bin is reached across many rounds. The route counts through a grid are exact arithmetic and are set out on the page about the edge bin and the middle bin, but turning those counts into frequencies requires assuming every route is equally likely — an assumption about the software that no seed check confirms and no operator in our table publishes.
It does not establish a return. No operator in our table publishes a return figure for any configuration, and the library has no field for one. A seed check produces a verified sequence of outcomes; it does not produce the number that would let two configurations be compared, which is why the page about row count and risk names no better setting.
Where a payout is actually decided
The seed procedure lives in the game engine. Every rule about money leaving the account lives in the terms, and none of the eight documents that publish a seed procedure connects the two.
Two of the eleven print an amount above which identity documents are demanded — Bitcasino.io at €2,500 in clause 6.6, Rocketpot at US$2,500 in clause 11.4 — and neither of those two is among the eight. Eight of the remaining nine publish discretion instead, set out under when the documents are asked for.
Three print a ceiling on how much may be withdrawn in a period, covered under withdrawal ceilings.
A round can be recomputed to the last digit and still sit behind a verification request and a weekly ceiling. Those are different documents, enforced at a different moment, by a different part of the operator.
What the tag is, as a piece of data
It is worth being precise about the status of the eight names above, because the tag looks like a verdict and is not one.
It records that the operator publishes such a procedure. Nobody at this site has run a seed check, holds an account, or has verified that any published implementation does what it describes.
That is the same standard the rest of the site holds to: a permit number is recorded as a number in a register entry, a threshold as a figure in a clause, and a seed procedure as a thing the operator publishes. Where a document could not be read, the cell says so. The full account of what was read and what was not is on the page about how these pages are put together.
Why it still matters which eight
Eight of eleven is a high rate compared with the library's 21 in 100, and that is not a coincidence: operators who build their own games have a reason to publish a way of checking them, since a house-built game has no outside studio's reputation standing behind it.
So the tag is best read as a fact about the operator's disclosure habits rather than about the game's behaviour.
An operator that publishes a seed procedure has published something checkable. An operator that also publishes a verification threshold and a withdrawal ceiling has published three checkable things. Only two of our eleven — Bitcasino.io and Rocketpot — publish both a threshold and a ceiling, and neither of them is one of the eight. Which is exactly why these columns are compared side by side on the main table rather than rolled into a single score.