A rating site's own database is not evidence — it is a table its owner can edit. Here is what we write to Arbitrum every day, what it makes impossible, and how to verify a single signal end to end in about five minutes.
The problem with any rating site, including this one
A site that scores signal channels keeps its numbers in a database, and the person who runs that database can change any row in it. Nothing on the page tells you whether a signal was recorded as it was published, quietly reworded three weeks later, or deleted because it embarrassed a channel that started paying.
That is not a hypothetical failure mode. It is the ordinary business model of the "top 10 signal channels" listings, and it is invisible from the outside: an edited row looks exactly like an honest one.
So the interesting question is not whether we promise to be honest. It is what we have made impossible for ourselves.
What actually goes on chain
When a signal is read from a channel, its text and the fields parsed out of it are written into a fixed byte format — a snapshot — and hashed with SHA256. Once a day every new snapshot goes into a Merkle tree, and the root of that tree, 32 bytes, is written to a contract on Arbitrum One.
The contract is 0xceacc40cf7b7c797d990dce7164fcccd87dad693. Each batch is keyed by its date, so a day can be written once and never rewritten: a second attempt at the same day is rejected by the contract itself, not by our own code deciding to behave.
As of 30 August 2026 that is 5,232 anchored snapshots in two roots — 5,174 in the first, which covered the whole history recorded before anchoring existed, and 58 in the first daily batch after it. Four signals recorded since this morning's root are not anchored yet; they go into tonight's. That gap is normal and it is the honest version of "within 24 hours".
Thirty-two bytes a day costs about a quarter of a cent. The price is worth stating because it explains why almost nobody does this: not the expense, just the absence of any reason to make your own records harder to edit.
What an anchor proves
Exactly two things.
A specific set of bytes existed no later than the moment its root was written. If the snapshot were rewritten afterwards, its hash would change, the root would no longer be reproducible from our rows, and the mismatch would be visible to anyone who checks — including to you, on the chain, without our cooperation.
A signal was one of a fixed, published set for that day. Removing an inconvenient trade from an anchored day breaks the tree in the same visible way. A channel's losing calls cannot be dropped after the fact to lift its hit rate.
Timestamping and non-removal. That is the whole claim.
What it does not prove
It does not prove a channel published that text. We wrote the snapshot down; the chain only witnessed when. What connects a snapshot to the live message is an ordinary, separate check: for 90 days after publication the source messages are re-read and compared with the stored snapshot, and every channel page carries a count of its signals whose text changed afterwards. Telegram's own "edited" mark is not used as evidence for this — it is set for formatting, buttons and link previews. In one batch of 100 messages carrying that mark, the number whose text had actually changed was zero.
It does not prove a trade was any good. An anchored signal that lost money is anchored just as firmly as one that won. The chain says nothing about quality, and a channel advertising "blockchain-verified" as if it meant "profitable" is selling you a category error.
It does not make us trustworthy. It makes one specific kind of dishonesty detectable. Everything else about a rating site still has to be earned the slow way.
How to check a single signal yourself
The point of anchoring is that this is checkable by a stranger who assumes we are lying.
- Open any signal on a channel page and follow its anchor mark, or start from the verification page. You get the exact bytes we recorded, their SHA256, the Merkle branch above them and the batch they belong to.
- Hash the bytes yourself. Any SHA256 tool gives the same digest, or ours is wrong.
- Walk the branch: hash your digest together with each sibling in order, and you end up with a 32-byte root.
- Read the root out of the contract on Arbiscan for that batch number, and compare. The chain's copy was written on a date the chain also records, and neither of those is ours to edit.
If step 4 disagrees with step 3, the record has been changed since it was anchored, and no explanation from us should satisfy you.
Why Arbitrum
Ethereum settles the record and Arbitrum makes writing to it cheap enough to do daily without the cost quietly becoming a reason to skip a day. A batch that is skipped when money is tight is not a guarantee, it is a habit.
The anchoring key holds a small balance and can do nothing except write roots: it cannot move funds, and the contract has no function that edits or removes a root once written — the owner included. A watchdog converts the wallet balance into remaining days at the current gas price and warns well before it runs out, because a proof system that silently stops proving is worse than none, and it fails in the direction that flatters us.
The part that is not technology
Anchoring answers one question: has this record been changed since it was written down. It says nothing about what gets recorded in the first place, which channels are indexed, or whether our scoring is fair. Those still rest on the methodology being public, on losses being shown rather than hidden, and on a rule that no listing tier moves a measured number.
What is different is that one class of promise stopped being a promise. You do not have to believe that the signals on this site were not edited after the fact. You can check.