Seed Rotation: What a Provably Fair Hash Check Cannot See

A provably fair check proves that a revealed seed matches the hash shown before play, so a round was not changed afterwards. It cannot see when seeds are switched, or why. If a game could choose the moment to rotate seeds based on incoming bets, every round would still pass the check. That gap is worth understanding.

What does a hash check prove?

In a provably fair game, the provider commits to a secret server seed by publishing its hash before play. After the seed is revealed, anyone can hash it and compare. A match proves the seed existed before the bets and was not changed afterwards. Our hash commitment check performs that comparison in your browser.

That guarantee is real, and narrow. A hash check answers one question about one seed. It is silent on others:

• When was this seed put into use, and when was it retired?
• Why did the game move to a new seed at that moment?
• Do results look the same early and late in a seed's life?

What is seed rotation?

A server seed is not used forever. At some point the game retires it, reveals it so past rounds can be checked, and starts using a new one whose hash was already published. That switch is a rotation. Rotations can be triggered by a player (for example, by changing their client seed), by a schedule, or by the system.

Rotation is normal and healthy: revealing old seeds is what makes past rounds checkable.

What is the theoretical attack?

The concern is not that a seed is changed after commitment. It is that the timing of a switch could be chosen.

In outline: suppose a system held several committed seeds in reserve and could compute, in advance, which one gives low early results. If it could switch to that seed just as a large bet arrived, the large bet would land on a low round. The new seed's hash was published beforehand, so every round still verifies. No hash check would ever fail.

Two things limit this in practice:

• Many shared-round games fix the whole sequence in advance. They derive every round from one long chain of hashes and publish the end of the chain before play. With the order fixed, there is nothing to switch.
• Player-supplied client seeds mix an input the server does not control into each result, which makes pre-computing outcomes much harder.

The attack is therefore most relevant to designs where seeds rotate per player or per session and the client seed is set by the system rather than the player.

What would it take to test for it?

A rotation question is a statistical one, about many rotations rather than one round. The data needed includes:

Most of this is not visible from public demos or result streams. Bet sizes in particular belong to the site that runs the game. That is why the Clash Watchdog lab does not publish rotation findings today: its published figures are counts of rounds, with sample sizes and intervals, and this question needs data it does not have. We describe it here so the limit of a hash check is clear.

What can a player do?

A few habits narrow the gap where the game allows them:

• Set your own client seed if the game lets you. An input you choose cannot be pre-computed by the server.
• Check a few revealed seeds against their hashes. It costs a minute and confirms the commitment was honoured. How to Verify a Provably Fair Hash Yourself walks through it.
• Do not read meaning into a run of low rounds after a seed change. Low runs are common in every crash game. The streak illusion explains why, and the patterns and hot hours check covers the related belief that some moments pay better.

Why does this matter beyond rotation?

Because it shows the general rule: a hash check is a floor, not a full guarantee. It proves a committed result was not changed. It does not prove the game returns what its provider states, and it does not prove every design choice around the commitment is neutral.

The return is a separate question, answered by counting many rounds and comparing the measured figure, with its interval, against the claim. That is the lab's core work for every game it covers, including provably fair ones such as Aviator and certified-RNG games such as Spaceman that cannot be checked round by round at all. For the open question of whether a game is rigged, see the is it rigged? check.

None of this makes any round predictable. A revealed seed explains a round after it is over. History describes behaviour. It does not predict the next round.