ChainRated vs other rating sites — what makes this different?

Two more differences have gone live since this thread was written, and one of them cuts against a story we would have liked to tell.

The record is timestamped outside our own database. Every day we take the snapshot of each signal exactly as it was published, build one hash tree out of them, and write the root into a contract on Arbitrum. 9 confirmed batches covering 10,569 signals are on chain today. What that buys you: if we ever quietly edited a signal after the fact to improve a channel's numbers — or our own — the edited text would no longer match a root that was published before the edit, and anyone can check that without asking us. Our database cannot prove itself; a public chain can at least prove when we said something. Verify a snapshot.

We re-read the original messages, and the popular theory did not hold. The common claim is that channels quietly delete losing calls. We now re-read published messages at the source and compare them with our stored snapshot: of 4,657 messages checked so far, 6 had their text changed after publication and 0 had been deleted. That is a boring result, and it is the one we publish. "Channels delete their losers" would have been a much better headline than "they mostly don't" — but the number is the number, and a site that only reports findings flattering to its own thesis is not measuring anything.

The third thing, already mentioned above and now quantified: 5 channels in the index are reposting one operation's signals, with a median lag of 1 to 11 seconds between the original and the copy. If you subscribed to three of them for diversification, you bought one feed three times.