A channel added to a catalogue arrives with months or years of earlier calls, the feed it chose to leave standing. What that history can and cannot show, what a blockchain anchor proves about it, and how to read a channel with the start of its live record in mind.
Where does a channel's record start?
On the day the channel was added. Every channel on ChainRated has a date from which its record is the site's own. Calls published after that date are read as they appear. Calls published before it were read from whatever the channel still had in its feed on that day.
Both kinds of call end up in the same table. Each is stored as a snapshot of its text at the moment it was read, and each is anchored on a public blockchain. What separates them is the time between publication and first read. For a call recorded live, that time is short. For a call published before the channel was added, it can be months or years, and during that time nobody outside the channel was keeping a copy.
This page calls the second kind imported history: calls published before their channel was added. The term describes timing and nothing else. An imported call can be accurate, and a channel with a long imported history can be an honest one. The question here is what such history can prove.
What can a feed from before that date not show?
It cannot show what was removed before it was read. A call deleted before the channel was added never reached the record. It is gone for every subscriber as well: Telegram's channels FAQ says a message deleted in a channel "will disappear for all subscribers", and a new subscriber sees the history that is left. What that does to a channel's total and its worst losing stretch is worked through in what a channel's feed looks like after the losing calls disappear.
Edits work the same way. A call edited before its first read is stored in its edited form, and nothing in the record marks that it once said something else. Calls are re-read at the source only while they are less than 90 days old. For an imported call that was already older than that when its channel was added, the copy taken on that day is the only reading there will be.
So for imported history, the record and the channel's feed agree by construction: both are the feed as the channel left it. Their agreement checks nothing. A record can only disagree with a feed for calls it read before the channel had a chance to change them.
The history may also belong to someone else. Telegram lets the owner of a group or channel pass ownership to another administrator. Usernames of public channels can be sold and transferred on Fragment, a platform for auctioning Telegram usernames. Open marketplaces list ready-made Telegram channels with the year they were created in the listing title, so age is sold as a feature. A feed that goes back four years shows that the channel existed for four years. Who posted those calls, and whether the same people run it today, the feed cannot say.
Finance has a name for history that arrives this way. Hedge funds that join a performance database can bring their earlier returns with them, and Philippe Jorion and Christopher Schwarz describe the result:
"Backfill bias arises when the fund's performance is not made public during some incubation period but then is added to the database presumably following good performance since the listing decision is voluntary." - Philippe Jorion and Christopher Schwarz, The Fix is In: Properly Backing out Backfill Bias, 2017
A working paper by Getmansky, Lee and Lo cites an estimate of that bias at 1.4% a year for one hedge-fund database between 1994 and 1998. A fund's earlier returns at least came from its own books. A channel's earlier calls are whatever was still in the feed.
How much of the record is history from before?
Most of it. As of 7 October 2026, the record holds 36,679 signals. Of those, 30,294, or 82.6%, were published before their channel was added, and 6,385 were recorded after. Of the 72 channels with signals, 70 have some history from before they were added.
How far back that history reaches varies a great deal. A channel's earliest call was published between 21 and 1,720 days before the channel was added, with a median of about 192 days. For a typical channel, more than six months of its record were written before anyone outside the channel kept a copy. In the widest case, the gap is more than four and a half years.
The share falls as live reading accumulates, because every day adds calls recorded as they appear. It rises whenever a channel is added with its past. A channel's page today rests mostly on calls no outside record captured at the time, which is why it helps to know which part of the record is which.
What does an anchor prove about an imported call?
That the call was in the database no later than the day its batch was anchored. The date the call carries comes from the channel.
The verification page states the limit in its own words: "An anchor proves that a specific set of bytes was in our database no later than the moment its root was written". It adds: "It does not prove a channel published that text".
This is how any trusted timestamp works. RFC 3161, the internet standard for timestamping services, describes the service as one that "supports assertions of proof that a datum existed before a particular time" (RFC 3161). A timestamp fixes a moment by which something already existed. It cannot reach back before the moment it was made, so an archive from 2023 anchored today proves that the archive existed today.
A channel page lists its latest replayed calls with the date each was posted. Hovering over a call's Anchored badge shows how many days after publication it was anchored and the number of its daily batch, and the batch's page gives the date of anchoring. For a call recorded live, the two dates sit close together, because new snapshots go into the next daily batch. For an imported call, the posting date comes from the channel and is stored inside the anchored bytes, while the anchor date is the day the call entered the database. The stretch between the two dates is the part of that call's life the anchor cannot vouch for.
The batch pages show what this looks like at scale. The batch anchored on 5 October 2026 holds 22,879 signal snapshots, while the batches on 4 and 6 October hold 78 and 222. A jump of that size is consistent with a large amount of earlier history entering the database at once, which is how imported history arrives. The batch page records the size and the time of anchoring. It does not record where each snapshot came from.
A long gap between the two dates has a second possible cause. Calls recorded before anchoring started at the end of August 2026 were anchored together in the first root, as described in how we anchor every signal on Arbitrum. Either way, a long gap marks a stretch the anchor does not cover.
What happens to calls older than 60 days?
They stay in the record and are anchored like every other call. Their replay against price history happens on request. A replay works out how a call would have ended from the price data after it was posted. Calls from the last 60 days are replayed in the regular daily run, which picks up a newly added channel's recent calls as well. Older history is replayed one channel at a time, when it is asked for, and no time is promised for it.
Until then, an older call can sit in the record without an outcome, and a missing outcome leaves the question open. Some of these calls are waiting for a replay. Others have incomplete or inconsistent levels, or a pair with no price history to replay against, and cannot be replayed at all. What a replay can establish even when it does run is set out in what a backtest can and cannot prove about a channel.
Two time windows on this page are easy to confuse. The 60 days concern replay: which calls get an outcome worked out in the regular daily run. The 90 days concern re-reading: which calls are checked against the channel for later edits and deletions. A call can be inside one window and outside the other.
Why not add the old and the new together?
Because the two parts carry different kinds of error, and a single number hides which part produced it. A short live record is noisy. With few calls, a wide range of results is plausible, and how many trades a hit rate needs before it means anything works through how wide. That noise shrinks as calls accumulate. Imported history can lean toward whatever the channel chose to keep in its feed, and adding more of the same history does not shrink a lean.
The investment industry's performance standard draws the same line. When a firm's track record has a break, the Global Investment Performance Standards (GIPS) require the periods before and after to be shown separately:
"The new or acquiring firm must not link performance prior to the break in the track record to the performance after the break in the track record." - CFA Institute, GIPS Standards for Firms 2020, provision 1.A.33
GIPS is written for investment firms and asset owners, and the rule is borrowed here as a habit of measurement: two parts of a record with different evidence behind them are read as two parts.
Old history is still evidence of a kind. The calls that survived can be compared with what the price did afterwards. Their style, frequency, leverage and type of setup are all visible. If the calls a channel chose to leave standing still lose money, that says something, because the channel had every chance to tidy them away. A clean and winning past from before the start date is also what a cleaned feed looks like.
Reading a channel with its start date in mind
The start date of the live record is the most useful single fact about a channel's history, and these steps put it to work:
- Find where the observation starts. On any tracker, ask from which date the record is its own. On ChainRated, the Anchored badge next to a call on the channel page says how long after publication it was anchored. A gap of weeks or months means the call entered the record after the fact, or was recorded before anchoring began.
- Look at the live part on its own. Count the calls recorded as they appeared and see how they ended. A few weeks of live record is thin evidence, and it is also the only part the channel could not shape after the fact. The checks in how to check a signal channel's track record in 10 minutes work on that part as well as on the whole.
- Treat imported history as one-sided. Losses in it count. A spotless past from before the start date fits an honest channel and a cleaned one equally well.
- Read a missing outcome as an open question. An older call without one may be waiting for its replay or may have nothing that can be replayed.
- Ask the anchor only for a date. It shows when a call was in the database at the latest. When the channel published it is a separate question.
A record kept from the day a channel was added cannot reach back past that day. What it adds is a part of the channel's history that the channel no longer controls, and that part grows with every call it posts.